January 30, 2025 • 47 mins
"We talk about the S-curve for technology and in the past we've said it can be 10 or 20 years before you reach the top. I think what's going to be really different about AI is that time frame is going to be massively compressed," Dr. Tony Wood, Chief Scientific Officer at GSK, explains to Bloomberg Intelligence in this episode of the Vanguards of Health Care podcast. Dr. Wood joins BI analyst Sam Fazeli to address the disconnect between market expectations for its future drugs in the context of GSK's R&D strategy and technology leverage. The company has five potential new drug approvals in 2025 and sees its early stage investments in AI and ML holding huge potential to catalyze future drug discovery and development.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:10):
Hello, and welcome to another episode of Vanguards of Healthcare.
I am Sam Fazzelli, Senior Farma analyst and head of
Global Industries Research at Bloomberg Intelligence. BI is part of
Bloomberg's in house research department. We cover over two thousand
equities and credits and have outlooks on more than nineteen
industries and one hundred market indicties and commodities. We published
(00:33):
deep dives on a large number of areas as well
as conductor a range of surveys such as attitude toward
working from home and as far as equity trading studies.
So what I wanted to do was to introduce a
conversation I'm going to have with Tony Wood. Tony is
the chief scientific officer of Glaxo Smith klin or GSK,
and we're going to have a conversation with a relatively
(00:54):
wide range, but a lot of it focused on the
application of technology to the conduct of biomedical research, particularly
from the perspective of a farmer company such as GSK.
So Tony, with that in mind, I just wondered if
you wanted to give us a quick introduction to yourself.
I do know that you have a background and you
were involved in the machine learning and AI implementations at GSK,
(01:16):
so please talk us through that for a few minutes.
Speaker 2 (01:19):
Yeah. Sure, So, first of all, Sam, Hi, great to
be here. Hello to your listeners. A little bit about
my background. I've been in the industry now for more
than thirty years, our last seven coming up on eight
of those with GSK, I described myself as a drug
(01:40):
hunter and a technologist. And I won't take a lot
of time on the call today explaining my background in detail.
But as you say, one of the things that I
have been accountable for since joining GSK in twenty seventeen
is building out their technologies capabilities. I did that along
side working in what we call the platform functions at
(02:04):
GSK prior to getting the job as chief Scientific Officer here.
Speaker 1 (02:09):
Okay, great, So let's start off with Obviously, any new
head of R and D that comes in, even though
you've been working at GSK for a certain length of
time or a long time, it will have their own vision.
So maybe you wanted to just touch on with us
what your overall vision for R and D at in
(02:30):
general and specifically at GSKA is compared to the prior strategy.
Speaker 2 (02:34):
Yeah, and I would say, let's start with how I'm
building on the prior strategy. You know, Hal and I
work together on putting in place a lot of the
technology that I'm going to talk about, but it's worth
noting some continuing changes that I've put in place since
since how moved on and I've been doing the job
(02:55):
now for almost two and a half years, and I
have three priorities for the group. The first one is execution,
and that's really all about the portfolio. I'll come on
and talk a little bit more about that in the moment.
And the next one is technology, and again i'll underscore
that for you in a few minutes too. And then lastly,
(03:16):
of course, neither of those two things get delivered without
the culture, the right culture, so we might talk a
little on that as well. Let's focus on the portfolio
first and really the changes that I've made there. If
you look at the overall performance of gsk R and
D against the sort of indices that are used across
(03:36):
the industry in benchmarking exercises, you find that in terms
of the number of new medicines launched and the costs
per launch over the past five years, we're pretty competitive
in fact, actually we're up a quartile. But if you
then look at the value per launch the BKA sales,
if you like, that's the area where we have more
(03:57):
work to do. We've been improving it over the past
few years, but we're still focused on doing more now.
One of the changes that I made in order to
drive that improvement and to put us on a footing
for a continuing trajectory, was to organize our R and
D efforts into therapeutic areas. And we have four therapeutic areas.
(04:20):
Those are infectious diseases, including vaccines, secondarily within that HIV
where we partner closely with THIEVE, and then respiratory immunology
and inflammation, and lastly oncology. And what I did was
appointed individual heads of each of those areas. And what
that allows us to do is to work very closely
(04:42):
with Luke and the commercial organization to ensure that our
project portfolio is pointed towards areas where we have both
significant medical need and epidemiology and scientific opportunity, and with
that comes an elevation in projected pks. So if you
look across the entirety of our portfolio now and we
(05:04):
have approximately seventy assets in development. You'll find that the
majority of those where targeting areas where PKS sales projections
are at least heading to what we call major blockbusters,
and that would be PKS sales of greater than two billion.
And then just quickly on technology, because I know you're
going to want to ask me more on this, Sam,
We've got to focus on four different areas. But broadly speaking,
(05:29):
we think about technology and two pieces data technology, which
is about understanding targets and patients, and platform technology, which
is about the design of medicines and vaccines to address
medical need once one has gained a better understanding of
a target and which patient would benefit from that. And
(05:49):
then lastly in technology, of course what we're focused on,
as indeed are many of our peers, are the improvement
of performance and things like clinical trials. I'll pause this, Simon.
I won't go on and talk about culture yet because
some I don't want this to turn into election.
Speaker 1 (06:06):
Thanks for that, Tony. Just in the context of portfolio
and the four therapeutic areas, So you have an area
that you've highlighted, which is mash and alcohol related fat
deliver disease with a new modality that you've discovered. How
do you deal with situations that don't necessarily fit the
four thas unless you tell me that actually does fit
(06:28):
one of the tas. But if you wanted to just
consider that in general.
Speaker 2 (06:31):
Yeah, Look, I mean we have we have something that
we call opportunity driven, but actually that's a diminishing part
of our portfolio. And in fact, the asset that you
describe is one that I would say is more about inflammation,
like getting increasingly interested in the long term consequences of inflammation,
(06:51):
particularly fibrosis, and the targeting question is something for which
the genetics are strongly thought of an impact in the
sort of later stages of inflammatory liver disease that are
where fibrosis begins. They're known as the F three and
four stages of the disease. And you can think about
(07:14):
in that particular context the inflammatory setup for that being
either as a result of something called stota stetosis, which
is fat driven inflammation, or in the case of alcoholic
liver disease, it's a consequence of both a steototic component
but also metabolic changes that occur as a result of
(07:38):
excess alcohol consumption. Now, this target itself has some interesting
features associated with it. As I say, it's very strongly
credential from a standpoint of human genetics. If your listeners
follow the area, it's almost PCSK nine like, so almost
the strongest validation that you can get. Interestingly, this is
(08:01):
a target that we're addressing with an ol nucleartide and
new platform technology. And the nice thing about that is
this particular target, although the genetics so well understood mechanistically,
it's sort of eluded us for a little while. But
the nice thing about all of those is you just
need simply to know in which direction do you need
to shift the expression of the target in order to
(08:23):
have an effect. It was very clear from this one
that reducing expression should have an impact. So that's some
I think, a really nice example of how the two
technologies come together. It's worth while saying in that case
as well that if you look at the detail of
how that target is expressed in different types of pathosites
(08:43):
using modern techniques like spatial transcriptomics and imaging, we have
a nice confluence, if you like, of all of the
factors adding up in the right direction, so that gives
us not only optimism for increased probability of success to
the target, but also points to new and additional targets
(09:03):
to follow it perfect.
Speaker 1 (09:05):
So if I take that to its obvious conclusion, I
shouldn't really even ask you whether you're interested in extending
that particular area into obesity, even though you could still
argue that obesity is an inflammatory disease. But let's not
do that. But I'm assuming therefore the glaxo I mean,
this can be a short answer yes or no. It
(09:26):
is not going to be expanding into that area.
Speaker 2 (09:29):
Look, it's super crowded, and obviously the jlp ones are
making an enormous impact on medicine. I've got an exciting
portfolio ahead of me, and as I mentioned, I'm more
focused on thinking about the future in which we might
be adding two glp ones to improve therapeutic effect in
(09:49):
the context of inflammation or indeed dealing with the residual
risk that is left once weight loss and was increasingly
an observation of ittional pharmacology associated with GLP one activity
has sort of run its.
Speaker 1 (10:06):
Course, right, So if we just go onto the idea
of developing the portfolio in general, there is obviously two
ways you could do that. You could do it organically.
Some of your assets are organic and you could do
it through BD. What should we be expecting Glackso in
terms of the to do going forward, in terms of
the do you have a set percentage or you opportunistick
(10:29):
with regards to the BED opportunities.
Speaker 2 (10:31):
No, Look, we have specific BD goals and that is
really in terms of an expectation that we will continue
to look for be the opportunity. I mean, look, if
you think about it, no matter how big an R
and D budget might be, it's still a fraction of
the total amount of resource that is spent in biomedical research.
(10:53):
I'm lucky in that the search function of BD reports
to me and we view by we, I mean Luke, myself,
David Redfern, and indeed all of the leaders in R
and D. We view our portfolio, whether it be discovered
internally or acquired through BD, through exactly the same lene.
(11:14):
So we treat programs coming into the portfolio in the
same way as you might imagined. There is a sophisticated
analysis of risk and value an opportunity for each of
those and it's very much part of our core that
we treat both inside and outside opportunities in the same way.
Speaker 1 (11:33):
Great, So if I just could dwell on that for
one moment, there is the history of Farmer has showed
us or suggested that a lot of beds done by champions.
There is a champion for a particular asset, and therefore
when heads of oncology or heads of neuroscience, etc. Move on,
suddenly that those assets become secondary. Has glacso managed to
(11:58):
change that buyer? Is it become? Are your systems in
such a way that once and as it comes in,
everyone has bought in and there is no champion left. Yeah?
Speaker 2 (12:07):
So well, and inevitably, you know, there are always champions
for individual assets when one finds them in the first place,
when you internalize them, and then they have formal project
leaderships leadership I expect those individuals to be, if you like,
the champions for their programs. But let me talk a
little bit through the process that we follow because in
(12:30):
each individual case, as I mentioned, we use the same
frame for judging internal versus external assets, and within that
there is an independently governed approach which would look at
the epidemiology, the market potential. We would access what we
call a probability of technical and regulatory success, and that
(12:50):
goes through individual expert teams and is then governed through
the same portfolio governance mechanism as I have for internal programs.
So to the greatest extent possible, the characterization of an
individual BD asset and the decision to prosecute it is
very much enshrined in that more quantitative approach. The championship
(13:15):
really comes in with regards to the energy and imagination,
if you like, we come with the design of plans
to accelerate and augment the value of individual assets.
Speaker 1 (13:26):
Great. So if I just have to touch on this,
it's an area where there's lots of discussion. I've been
doing some digging into the science space in China and
you can see there are there by any measure and
Nature does quite a good assessment of percentage of top
one percent citations, etc. And you can see that there
(13:49):
is a hockey stick effect going on there. But so
there's this discussion that there's been lots of licensing deals
from zero in twenty nineteen, let's say that's correct to
a particularly high number I have we're just finish finalizing
our assessment of the details, but several tons of deals
and you've done some yourselves, particularly ADC areas or t
(14:10):
sell engagers, et cetera. And so there's two things I
wanted to ask you. Number one is do you what
do you think has changed in your interactions with Chinese
biotaics that has enabled this surge. It's as simple as
the investment in science. And then do you see a
risk that some way politics will get in the way
(14:32):
of this, just as it's such a situation such as
the Biosecures Act.
Speaker 2 (14:37):
Yeah, And like Sam, I would say, probably the first
thing to say about m BD is obviously we have
a team that is global and we look to identify
innovation that is right for GSK wherever in the world
the science with the right characteristics emerges, and it would
be right to say that we're currently seeing an increase
(14:59):
in in volume of innovative science coming out of China.
As you say, we have a number of partnerships ourselves
and look in the context of managing the geopolitical situation,
Obviously we're very alert to whatever legislation is developing, and
we take account of that when we're making decisions about
(15:21):
individual assets, particularly with regards to the need to consider
ultimately the global manufacturing footprint that we have for one
asset versus another. So I would say this is just
simply part of the emerging and changing landscape of global science,
and we have the right mechanisms in place to spot
(15:42):
new areas of innovation and also to manage across the
legislative framework.
Speaker 1 (15:47):
Okay, so let me come to a particular asset, a
resurgent asset within the GSK portfolio, which is blend rep.
And the reason I mentioned this under our strategy and
broad discuss is clearly the first time round there was
something that was missed in the R and D program
in terms of the way that maybe do seeing whatever
(16:10):
the attitude was that ended up getting us to a
point where the side effect issues and then the trial
as a single agent didn't work out, et cetera. And
then of course now we have dreams seven and dream
made with very strong overall survival benefit. Just for the record,
we have a very deep analysis of multiple maloma and
we are big fans of blend rep. We think that
(16:32):
is a particularly well placed asset in the armamentarium of
treating a multiple myloma. So what do you think what
the lessons that any R and D organization And of
course in your case, yours can learn from the process
which blend Rep went through.
Speaker 2 (16:47):
Yeah, and look, let me just sort of reflect on
a few of the points. So you've made sam and
for the benefit of your listeners. Blend initially had conditional
approval in late line my loma, and in conditional approval
cases one has to have a confirmatory study, and that
(17:09):
for blen Rep was Dream three and that was a
monotherapy blend Rep versus comdex standard of care combination therapy,
and it was in late line. And when we look
at that, I think clearly, although blen rep has interestingly
a number of mechanisms of action, the decision to compare
(17:30):
monotherapy with standard of care is something that was on
behind the ultimate Dream three failure. And also that we
were looking for superiority versus standard of care. So I
think we just sort of shot for the stars in
too great an extent. In the Dream three study, of course,
we had Dream seven and Dreammate going on as well
(17:51):
whilst that particular study was being conducted, and Dream seven
and Dreammate were in the second line setting, so less patients,
and they were blend Rep as part of standard of care.
So if you like a more realistic comparison, and as
you mentioned, you know what we got from that, fortunately
(18:13):
was a fantastic result in terms of efficacy with blend
rep one, tripling the progression free survival for patients in
second line and reducing the risk of death even so
by forty two percent. And it's really unusual in MYA
loma to have both pfs and OS and this is
statistically significant OS. So we reported at the end of
(18:36):
last year. What we were also able to do through
the Dream seven Dreammates and indeed other clinical programs was
gain a better understanding of the ocular side effects of
the medicine. Dream three was the first time that those
side effects became apparent, and of course, over time what
one us learns more effectively how to deal with them.
(18:57):
And what we know now about the the blur vision
is that it's some reversible, transient and manageable and in fact,
in the vast majority of patients on studies, it's possible
to reverse the effect by changing the treatment schedule. And
if I just refer back to Dream seven, what we
(19:19):
know there is that more than seventy percent of the
patients had some form of schedule modification to manage ocular
side effects, and we still had those fantastic efficacy results
that I discussed earlier. So I think it's really just
about learning about a new medicine. You'll appreciate that learning
(19:40):
to manage the side effects of new cancer measure medicines
is not something new. You know, there are a number
of cancer medicines with significant side effects interstitial lung disease,
cardiovascular effects. So the cancer physician community are very used
to managing side effects. As they say in this particular instance,
(20:02):
that they're reversible and transient and very manageable.
Speaker 1 (20:07):
Right, great, yep. I have to say I agree with
most of that, given all the conversations we have with
physicians and various surveys we've done. So let's just to
wrap up on the broader strategy. What is Is this
something you can point to that's different about GSK's R
and D focus or R and D strategy to peers.
(20:29):
I mean, there are lots of peers, so it's unlikely
that there is any significant difference, But is this something
you wanted to point out before we moved on to
other topics.
Speaker 2 (20:39):
Yeah, and I won't repeat the focus across individual therapeutic areas,
but I think for us it's very much. We talk
about the approach of science times technology times culture, and
it really is the embedding of technologies, particularly in our
early stage portfolio, but also bling development that I think
(21:02):
sets us apart, and a culture as we touched on
earlier as well, that welcomes innovation both internally and externally.
And the point that I made with regards to external
partnerships and later stage assets also applies to our earlier
stage portfolio. You've no doubt quollowed that we had a
range of early stage collaboration deals that were announced towards
(21:25):
the end of last year as well, things like Relation
Therapeutics OPRAH or the Salius and we can talk about
each of those, but at their heart, these are biotech
organizations who have the same approach to technologies we do
at GSK. That is, the underpinnings of target and patient
choice coming from human genetics, and then the reinforcement of
(21:49):
those observations in what we might call the phenotypic features
of disease. I think those two things together give a
unique insight into choosing the right target and matching them
to the right patients at the right stage of disease.
Speaker 1 (22:03):
So would you say last question on the strategy R
and D et cetera. Would you say that that is
where the market underestimates GSK's R and D approach or capabilities,
not not specific assets, because we can talk about those
consensus versus company and I think at the JP Morgan
conference and at the Early Stage session that he had
(22:24):
just before Christmas, you highlighted the whole list of those.
But in terms of the broad potential for GSK's R
and D, is that where most people are underestimating or
is there another area you would like or areas that
you would like to highlight?
Speaker 2 (22:37):
No, would I would just say the look in general
consensus value is really focused on the parts of the
portfolio that the market can accurately model, and we might
talk about differences across particular assets. We've covered one indeed
with blend Rep, where I think the market is likely
(22:58):
waiting to see the nature of the lake that we
get with regards to REMS. And that's understandable given the
history of the asset that you've just described. They don't
tend to assign a great deal of value to early
stage technology, and it's why programs like for me, the
one that we started our conversation with the arrow eight
(23:20):
THESK ninety ninety molecule, that should we be successful with
a phase two proof of concept there that I can
point to something that not only brings value MASH and
alcoholic liver disease in particular has significant epidemiology. I think
it's the largest cause of liver transplantation worldwide, and if
(23:42):
I remember correctly, something like twenty six million individuals with
some form of alcoholic liver disease for example. So if
we can show with programs like that that we have
the marriage of scientific opportunity through genetics and function genomics
and the application of new platform technologies like all nucleotides,
(24:06):
and to marry that to opportunity, then I would hope
with increasing examples like that the market we'll see that
our approach is one that is indeed reducing risk and
enabling us to identify significant areas of epidemiology scientific opportunity.
Speaker 1 (24:23):
And from that value, I think this is a perfect
bridge to talking about technology and AI applications, et cetera.
So just let's ask you one hopefully very straight simple question.
There's nothing simple in this world. I guess, where do
you think we are in our ability to extract value
from AI and mL under curve. Would you say we're
(24:46):
in the first innings, whichever metaphor you want to use.
Speaker 2 (24:50):
Look, I'd say we're very much just at the bottom
of that S curve. And you know, we talk about
the S curve for technology and in the past we've
said it can be ten or twenty years before you
reach the top. I think what's going to be really
different about AI is that timeframe is going to be
massively compressed. Ultimately, though, I think what is really important
(25:11):
is to understand that the AI is really only as
good as the data that informs it, and that certainly
underpins our strategy. But I would say they're already and
we're seeing it internally as well. Areas where AI is
having a significant impact on the efficiency of what we do.
The design of small molecules, for example, is an area
(25:33):
in point, which is one that has been pursued both
internally in GSK and externally, and again that's a reflection
of where people can find high quality data. Protein structure
designed to go back to alpha fold, the PDB was
a source of information for the derivation of alpha fold.
(25:55):
The use of generative AI for the construction of regulatory
documents that I think is is already proving to be
an area of great advantage and efficiency. And again it's
one that we are also pursuing.
Speaker 1 (26:12):
Can I just stop you there, because you just mentioned
quite a few things. Let's take them one at a time,
and I'd love your perspective in this. Let's start with
the one that's probably the and you shoot me down
if you want, because I might be just talking rubbish here.
But preparation of regulatory documents is that one of the
low hanging fruits with the use of generative AI.
Speaker 2 (26:34):
Yes, I think it is. I mean clearly, in anything
you do with the regulator, one has to approach that
in a careful fashion and in partnership. But these documents,
wherever you can see an increase in quality and being
able to remove human intervention helps to improve the quality
(26:54):
of those documents. So that is something that works very well.
Of course, what it requires is the flow of information
into the document in the first instance, as well as
you know, the authoring of the document, if you like,
is merely the last stage in a complicated process. But
it is proving to be an effective approach. It obviously
(27:16):
requires human intervention at the end, because the thing still
may make errors, but it is improving efficiency enormously. I'm
more excited in clinical trials setting about some of the
other opportunities that AI.
Speaker 1 (27:32):
Right, right, let's let's let's come back to that in
a second. But I wanted to to see if you
would put a number on it. Tomorrow, say dream nine
top line is announced with the help of AI, how
much faster can you get the data to regulators than
you would have done without it?
Speaker 2 (27:51):
Yeah, and I don't think I've shared that data externally.
Sounds so what I'm not?
Speaker 1 (27:56):
But what's your What would you say if I put
thirteen heads of R and D the rough number would
be not necessarily GSK.
Speaker 2 (28:04):
I'd say, you know, they'll all substantially say that this
could come down to a matter of perhaps weeks. What
I would say for the final preparation of the documents,
what I would say internally is that when we look
again in our performance with comparators, then we're very well placed.
(28:25):
In fact, I think in the last CMR analysis we
were sort of second in the large Farmer group in
terms of that stage of the regulatory process. But as
I say, I think there are much much more exciting opportunities.
Speaker 1 (28:38):
There, right, So let's step back from that, because that's
the Actually, there's two sides we can go. We can
go forward and look at the patient identification and making
sure that the drugs get to the right physicians and
patients best educated fast as possible. So that's the launch
side or the step before it, whichever one you to
(29:00):
talk about next, which is the actual conduct, patient recruitment,
and the conduct of the trials.
Speaker 2 (29:06):
I'll take it in pieces, because let's let's do that next.
Patient conduct and recruitment of trials, the performance of individual sites.
This is an area where I think it is going
to be it already is in fact possible to improve
clinical trial recruitment. Behind that as well, there's an increasing
(29:26):
ability to examine a wealth of data that we might
refer to as secondary endpoint that determine, if you like,
the potential trajectory or risk factors for individuals, not necessarily
always encountered in the in the clinical trial design up
(29:47):
to now, but an ability you know what, one of
the things we often find in our in our studies
is we have a statistical hierarchy for analysis that requires
that you have points listed in such a way that
you have to succeed in each of them in order
to submit the subsequent one into a label. I think
(30:08):
AI is going to be very helpful in ensuring that
we understand how individual if you like, heterogeneity across broader
patient populations impact those types of outcomes. It's also interesting
to note that for a lot of our baseline characteristics,
people are judged at the value at which they are
entered into the study, and yet those characteristics themselves are
(30:31):
also not stable, and so you're looking at a trajectory,
and AI is doing a much better job increasingly beginning
to be able to project those trajectories. So there are
a host of different ways in which one might imagine
not only improving the recruitment through the performance of individual sites,
if you like, but also improving the characteristics the power
(30:53):
of your studies by better understanding the likely behavior of
patients and control arms. There's a lot more to be
done on this SAM, but I think you can already
see some of the opportunities that are emerging.
Speaker 1 (31:07):
Can AI replace the DSMBs in a live fashion if
AI is cognizant of the data that's coming out on
a daily, hourly basis, from the trials. Does that constitute
a problem for the regulator if no one has got
a view on that data?
Speaker 2 (31:24):
Well, no, I would say. In all of this, what
AI is doing is positioning those decision making bodies in
a better place to make higher quality decisions, more effectively
and indeed more quickly. So I don't see it as
replacing any of the independent decision making bodies. It simply creates,
I think, a higher quality proposition for them.
Speaker 1 (31:46):
And what are the issues? To use a phrase that
I think a lot of people hate to use, but
it was a good one, is the unknown unknowns? Is
AI better at connecting the dots in areas where humans
perhaps have bias and therefore don't look for it? So,
you know, the example being we always look for our
keys under the light. So do you think that could
help in terms of identifying issues or opportunities earlier as
(32:10):
the data flows.
Speaker 2 (32:11):
Yeah, well, look, I mean, let me step back from
the clinical trial environment and talking about this, because I
let's go first of all to where I think in
the early stages of the R and D process, AIML
again is showing its value already, and that is, to
your point, the ability to connect across vast data sets
(32:35):
near one level. Let's consider, for example, the use of images,
which are becoming increasingly prevalent as endpoints in early stage research.
I've been amazed by how good image analysis is at
getting at underlying elements of pharmacology. Now you need an
ability to qualify the components of the image that the
(33:00):
AIML is using to highlight changes. You can we call
them latent space variables, because an image contains a huge
amount of data, and so if you like separating signal
to noise, requires that you have the tools in order
to do that. But once you've been able to establish,
(33:21):
if you like, an abstraction of the image, which is
a faithful representation of the pharmacology that you're looking for,
then it is tremendously effective at finding features and indeed
potentially at identifying aspects of farm molecule performance that you
may have not initially spotted using a more specific endpoint.
(33:44):
Of course, in all of those cases you have the
reassurance that whatever AIML suggests, you can qualify your answer
by running an experiment. And that's why I went to
that area of research first. In the clinical arena, of course,
one has to be very very much more aware of
the potential risks associated with the use of an AI approach,
(34:08):
and of course there is emerging regulation there and we
remain very aligned to that approach and ensuring patient safety.
Speaker 1 (34:17):
So what do you think would be the poster child
of the first asset that is moving forward within GSK
that has leveraged all of these areas did genomics. I've
got a feeling you might say it's the alcoholic liver
disease or asset.
Speaker 2 (34:32):
But well, ultimately, if we get what I would call
clinical proof of concept, we have early results showing that
that molecule reduces the transcript for the targeting question and
that we reduce ALT, which is, as I'm sure you're aware,
a marker of liver injury. Perhaps we might come back
and talk about it next year and I see what
(34:53):
it does deliver fibrosis. But an early example in the
portfolio where we view AIML would be our treatment for
chronic hepatitis B for example. This is an interesting area
because in order to achieve what's called functional cure, one
needs to reduce the level of viral DNA and to
(35:18):
reduce a surface mark or tent called the surface antigen
to a level which is below quantification. Now for our
particular molecule, this is another Oligan nucleotide as it happens
at the reversant, it's the only molecule that is able
to do that in patient to a significant effect in patients.
(35:40):
We showed this in phase two studies where we had
a greater than fifteen percent reduction of both of those
features below quantifiable levels. Our models, though, which are based
on not only those factors but other liver measurements, show
that you have to get much lower than than the
quantum fiable limits to see a functional effect. And what
(36:03):
we were also interested in was the interplay between immune
system changes and the if you like, more traditional virology.
And what we've been able to do there is to
use a very is to use an approach with AIML
and imaging. On top of those are the quantitative measures
and that we feel is going to give us an
opportunity to perhaps help up to fifteen percent more patients
(36:25):
than just using the traditional approaches. So, and you know,
I could mention other areas that we're developing, but that's
the general theme that you will see come through that
we're able to better qualify patient response by a combination
of both image related data and complex molecular markers.
Speaker 1 (36:46):
Great. So, just to summarize all this, if I were
to look to ask you where the biggest impact will be,
Let's say you have the percentage that you wanted to
put on it, and it has to add up to
one hundred percent improved probablity of success, faster time to market,
and faster speed to reach peak.
Speaker 2 (37:04):
Yeah, man, let me sort of. And I think, of
course a lot of that depends on what point in
time are you going to judge it. I think, in
the first instance, what we'll see the efficiency gains, so
faster time from study to file, for example, you also
see efficiency gains in the discovery of the initial molecular
(37:28):
entities in the first place, small molecules or proteins. GENAI
works there, and it's not unusual to expect that you
might see our perhaps and I'm taking a very ball
part for you here, rather than our own internal data
that being reduced by as much as some fifty percent,
if not more. I think in time, though, what we'll
(37:49):
see is an improvement in our ability to match molecules
and targets to patients. And that's really what the important
goal is. You're a preciate that our Phase two success
rates across the industry are still relatively low. I mean
typically it's a one in twenty success. We talk about
(38:10):
the fact that genetics might improve that by twofold. Interestingly,
some of the work that we're doing looking at single
cell phenotyping, which is where this ability to use AIML
to layer different levels of data come in, they also
suggest that you might see perhaps a two or threefold
improvement in phase one. So I think ultimately the big improvements,
(38:33):
and name these may be in the order of a
fivefold increase in success will come in phase two studies,
where we typically at the moment test whether or not
our initial hypothesis was correct or not great.
Speaker 1 (38:47):
So I think I've answered every element that I just highlighted.
So is the implementation of AI and mL going to
be absorbable save by the industry as a whole within
current R and D budgets or will we have to
have a rise before we see the benefits?
Speaker 2 (39:05):
Well, at least the approach I can I can speak
to for GSKS were very much absorbing this within the
context of our existing R and D budget, And you know,
the change requires as any sort of change program does that.
I've always found that it's not only a matter of
adopting the new, but it's a matter of stopping the old,
(39:26):
and that that gives you an opportunity then to realize
these these changes. The big the big change for me,
that is the generation of more relevant, if you like,
human subject based data, and as I mentioned earlier, you
can see that in some of the collaborations that we've
put in place towards the end of last year, as
well as a more longer lived investments we've made in
(39:49):
human genetics. So all of that I think allows us
to make more effective decisions early on through the use
of highly autom mated systems, so we get to analyze,
if you like, more potential programs more rapidly, and then
in terms of the shape of the funnel, to be
(40:10):
able to pick ones that are going to succeed with
higher likely subsequent probabilities. Means that when you get into
the clinic, then one is able to run perhaps a
slightly smaller early stage clinical profile portfolio sorry with greater confidence.
That's where ultimately I think the change in the approach
(40:31):
that we will take will be most reflected in the
overall characteristics of R and D performance.
Speaker 1 (40:39):
So totally I'm conscious of the time. So I want
to help us wrap up here. Of course, I'll give
you the last word afterwards if you want to say
something that I haven't asked. But in the world of drugs,
in the years that I've been following it and analyzing it,
we've had darling areas or areas where they've been they've
become multi blockbus areas. So you start it off with
(41:00):
I think an area that GSK was involved in themselves,
the antacit world time. Then of course we moved on
to the to the lipid lower engagents, and then came
the era of anti tnfs and equivalents, and then obesity
now and is Glaxo do you believe invested in one
of the therapeutic areas that will become the next big
(41:22):
whatever you want to call it multi blockbuster and which
one is it?
Speaker 2 (41:26):
Yes? So I'm bound to say yes, aren't i? Sam?
And look, I mean the only one I would add is,
of course the PD one io in cancer.
Speaker 1 (41:35):
Of course, yes, sorry I missed that one.
Speaker 2 (41:37):
The characteristics of those areas as they tend to be
the culmination of high medical need or epidemiology, a new
understanding of biology, and often a new medality. Right you
You you self described the journey from small molecules into
into monoclonals in your in your summary of progress. I
(41:59):
would hope now that with some of the tools that
we've just described, we'll be better able to forecast those
areas than we have done in the past, because the
one thing I would say about that history is that
it's not usually the same organization that finds success. The
industry as a whole has succeeded, but individual organizations have
(42:20):
not repeated that. I would point to perhaps two broad
areas that where laying the foundations. I would say in
beginning to experiment in the first one, we've already covered,
and that's some tissue fibrosis following inflammation. I think somewhere
in the region of almost forty percent of deaths teaches
(42:40):
some form of fibrotic organ dysfunction or failure. And you
could almost within that include neurodegeneration. Where we're beginning, because
I think the mechanistic pathway to newer degenerative diseases is similar.
It's inflammatory and in nature and ultimately involves the consequences
(43:06):
then of that on loss of brain material and function.
We're beginning to invest in that area. We have collaborations
with Elector looking at progranulin modification. This points to a
very common area of biology and neurodegeneration.
Speaker 1 (43:25):
Is that often talked about that particular relationship.
Speaker 2 (43:29):
No, well, this is lyce'somal dysfunction, which I think does
is definitely mechanistic of mechanistic importance. We have obviously and
we are hopefully be able to say more about this soon.
Ongoing collaborations with health authorities looking to design not typical
(43:50):
perspective face three studies, but nevertheless perspective studies to evaluate
whether or not shingles vaccination protects from neuregenerator.
Speaker 1 (43:58):
I took one as soon as I said, read paper.
Speaker 2 (44:02):
And so those sorts of areas you'll see us continue
to experiment in. And as I mentioned earlier, the general
realization that the journey from autoimmune disease, which of course
is a cause of inflammation in and of itself, into
examining fibrotics sequally is something that our portfolio will begin
to take a closer look at. You can see the
(44:25):
beginnings of that in our work in lung disease in
liver disease, and we have some earlier programs in in
other end organs. We haven't disclosed them externally, so I'll
leave it at that. But there's a general theme here
of emerging new biology behind fibrosis and a new approach
to the newer degeneration.
Speaker 1 (44:47):
Did you I want to let you finish off on
any topic that you wanted to wrap up on. If not,
I have one last question and we can. It's a
very simple yes.
Speaker 2 (44:55):
No, And I think you've done a good job of
a journey through.
Speaker 1 (44:57):
Okay, twenty twenty five. Will we have an opportunity to
see IRA equalizes between small molecules and biologics in terms
of negotiation time?
Speaker 2 (45:06):
Oh? Look, I mean I'd say that is a matter
of discussions with the new governance in the US, and
all I would say there is Look, we've worked with
the Trump government in the past, we work with them again.
US continues to be an important area of innovation for US.
We've just put eight hundred million pack dollars into a
(45:27):
new film finish capabilities. So I generally tend to spend
my time focused on the portfolio. I account for the
various legislative differences that appear and and honestly, I think
you know, we still find a path to exciting programs
within whatever context that we have there. And probably the
(45:49):
thing I would say to finish is just on that
that I'm looking forward to another exciting year here. We
had thirteen successful Phase three studies last year, and that
of course ECHO is out now into five important launchers
heading our way, not the least of which, of course,
with a producer date in the middle of the area
is is blend rep in second line my loma. So
(46:12):
really looking forward to seeing how that works out. And
I've enjoyed our call today. Pleasure to spend time talking
to you.
Speaker 1 (46:20):
Thank you very much, Tony Wood, chief Scientific Officer at
GLEXO GSK. You have to get out of the habit
of saying that, thank you very much for spending the
time with US user US and
Speaker 2 (47:06):
USAGES
Hello, and welcome to another episode of Vanguards of Healthcare.
I am Sam Fazzelli, Senior Farma analyst and head of
Global Industries Research at Bloomberg Intelligence. BI is part of
Bloomberg's in house research department. We cover over two thousand
equities and credits and have outlooks on more than nineteen
industries and one hundred market indicties and commodities. We published
(00:33):
deep dives on a large number of areas as well
as conductor a range of surveys such as attitude toward
working from home and as far as equity trading studies.
So what I wanted to do was to introduce a
conversation I'm going to have with Tony Wood. Tony is
the chief scientific officer of Glaxo Smith klin or GSK,
and we're going to have a conversation with a relatively
(00:54):
wide range, but a lot of it focused on the
application of technology to the conduct of biomedical research, particularly
from the perspective of a farmer company such as GSK.
So Tony, with that in mind, I just wondered if
you wanted to give us a quick introduction to yourself.
I do know that you have a background and you
were involved in the machine learning and AI implementations at GSK,
(01:16):
so please talk us through that for a few minutes.
Speaker 2 (01:19):
Yeah. Sure, So, first of all, Sam, Hi, great to
be here. Hello to your listeners. A little bit about
my background. I've been in the industry now for more
than thirty years, our last seven coming up on eight
of those with GSK, I described myself as a drug
(01:40):
hunter and a technologist. And I won't take a lot
of time on the call today explaining my background in detail.
But as you say, one of the things that I
have been accountable for since joining GSK in twenty seventeen
is building out their technologies capabilities. I did that along
side working in what we call the platform functions at
(02:04):
GSK prior to getting the job as chief Scientific Officer here.
Speaker 1 (02:09):
Okay, great, So let's start off with Obviously, any new
head of R and D that comes in, even though
you've been working at GSK for a certain length of
time or a long time, it will have their own vision.
So maybe you wanted to just touch on with us
what your overall vision for R and D at in
(02:30):
general and specifically at GSKA is compared to the prior strategy.
Speaker 2 (02:34):
Yeah, and I would say, let's start with how I'm
building on the prior strategy. You know, Hal and I
work together on putting in place a lot of the
technology that I'm going to talk about, but it's worth
noting some continuing changes that I've put in place since
since how moved on and I've been doing the job
(02:55):
now for almost two and a half years, and I
have three priorities for the group. The first one is execution,
and that's really all about the portfolio. I'll come on
and talk a little bit more about that in the moment.
And the next one is technology, and again i'll underscore
that for you in a few minutes too. And then lastly,
(03:16):
of course, neither of those two things get delivered without
the culture, the right culture, so we might talk a
little on that as well. Let's focus on the portfolio
first and really the changes that I've made there. If
you look at the overall performance of gsk R and
D against the sort of indices that are used across
(03:36):
the industry in benchmarking exercises, you find that in terms
of the number of new medicines launched and the costs
per launch over the past five years, we're pretty competitive
in fact, actually we're up a quartile. But if you
then look at the value per launch the BKA sales,
if you like, that's the area where we have more
(03:57):
work to do. We've been improving it over the past
few years, but we're still focused on doing more now.
One of the changes that I made in order to
drive that improvement and to put us on a footing
for a continuing trajectory, was to organize our R and
D efforts into therapeutic areas. And we have four therapeutic areas.
(04:20):
Those are infectious diseases, including vaccines, secondarily within that HIV
where we partner closely with THIEVE, and then respiratory immunology
and inflammation, and lastly oncology. And what I did was
appointed individual heads of each of those areas. And what
that allows us to do is to work very closely
(04:42):
with Luke and the commercial organization to ensure that our
project portfolio is pointed towards areas where we have both
significant medical need and epidemiology and scientific opportunity, and with
that comes an elevation in projected pks. So if you
look across the entirety of our portfolio now and we
(05:04):
have approximately seventy assets in development. You'll find that the
majority of those where targeting areas where PKS sales projections
are at least heading to what we call major blockbusters,
and that would be PKS sales of greater than two billion.
And then just quickly on technology, because I know you're
going to want to ask me more on this, Sam,
We've got to focus on four different areas. But broadly speaking,
(05:29):
we think about technology and two pieces data technology, which
is about understanding targets and patients, and platform technology, which
is about the design of medicines and vaccines to address
medical need once one has gained a better understanding of
a target and which patient would benefit from that. And
(05:49):
then lastly in technology, of course what we're focused on,
as indeed are many of our peers, are the improvement
of performance and things like clinical trials. I'll pause this, Simon.
I won't go on and talk about culture yet because
some I don't want this to turn into election.
Speaker 1 (06:06):
Thanks for that, Tony. Just in the context of portfolio
and the four therapeutic areas, So you have an area
that you've highlighted, which is mash and alcohol related fat
deliver disease with a new modality that you've discovered. How
do you deal with situations that don't necessarily fit the
four thas unless you tell me that actually does fit
(06:28):
one of the tas. But if you wanted to just
consider that in general.
Speaker 2 (06:31):
Yeah, Look, I mean we have we have something that
we call opportunity driven, but actually that's a diminishing part
of our portfolio. And in fact, the asset that you
describe is one that I would say is more about inflammation,
like getting increasingly interested in the long term consequences of inflammation,
(06:51):
particularly fibrosis, and the targeting question is something for which
the genetics are strongly thought of an impact in the
sort of later stages of inflammatory liver disease that are
where fibrosis begins. They're known as the F three and
four stages of the disease. And you can think about
(07:14):
in that particular context the inflammatory setup for that being
either as a result of something called stota stetosis, which
is fat driven inflammation, or in the case of alcoholic
liver disease, it's a consequence of both a steototic component
but also metabolic changes that occur as a result of
(07:38):
excess alcohol consumption. Now, this target itself has some interesting
features associated with it. As I say, it's very strongly
credential from a standpoint of human genetics. If your listeners
follow the area, it's almost PCSK nine like, so almost
the strongest validation that you can get. Interestingly, this is
(08:01):
a target that we're addressing with an ol nucleartide and
new platform technology. And the nice thing about that is
this particular target, although the genetics so well understood mechanistically,
it's sort of eluded us for a little while. But
the nice thing about all of those is you just
need simply to know in which direction do you need
to shift the expression of the target in order to
(08:23):
have an effect. It was very clear from this one
that reducing expression should have an impact. So that's some
I think, a really nice example of how the two
technologies come together. It's worth while saying in that case
as well that if you look at the detail of
how that target is expressed in different types of pathosites
(08:43):
using modern techniques like spatial transcriptomics and imaging, we have
a nice confluence, if you like, of all of the
factors adding up in the right direction, so that gives
us not only optimism for increased probability of success to
the target, but also points to new and additional targets
(09:03):
to follow it perfect.
Speaker 1 (09:05):
So if I take that to its obvious conclusion, I
shouldn't really even ask you whether you're interested in extending
that particular area into obesity, even though you could still
argue that obesity is an inflammatory disease. But let's not
do that. But I'm assuming therefore the glaxo I mean,
this can be a short answer yes or no. It
(09:26):
is not going to be expanding into that area.
Speaker 2 (09:29):
Look, it's super crowded, and obviously the jlp ones are
making an enormous impact on medicine. I've got an exciting
portfolio ahead of me, and as I mentioned, I'm more
focused on thinking about the future in which we might
be adding two glp ones to improve therapeutic effect in
(09:49):
the context of inflammation or indeed dealing with the residual
risk that is left once weight loss and was increasingly
an observation of ittional pharmacology associated with GLP one activity
has sort of run its.
Speaker 1 (10:06):
Course, right, So if we just go onto the idea
of developing the portfolio in general, there is obviously two
ways you could do that. You could do it organically.
Some of your assets are organic and you could do
it through BD. What should we be expecting Glackso in
terms of the to do going forward, in terms of
the do you have a set percentage or you opportunistick
(10:29):
with regards to the BED opportunities.
Speaker 2 (10:31):
No, Look, we have specific BD goals and that is
really in terms of an expectation that we will continue
to look for be the opportunity. I mean, look, if
you think about it, no matter how big an R
and D budget might be, it's still a fraction of
the total amount of resource that is spent in biomedical research.
(10:53):
I'm lucky in that the search function of BD reports
to me and we view by we, I mean Luke, myself,
David Redfern, and indeed all of the leaders in R
and D. We view our portfolio, whether it be discovered
internally or acquired through BD, through exactly the same lene.
(11:14):
So we treat programs coming into the portfolio in the
same way as you might imagined. There is a sophisticated
analysis of risk and value an opportunity for each of
those and it's very much part of our core that
we treat both inside and outside opportunities in the same way.
Speaker 1 (11:33):
Great, So if I just could dwell on that for
one moment, there is the history of Farmer has showed
us or suggested that a lot of beds done by champions.
There is a champion for a particular asset, and therefore
when heads of oncology or heads of neuroscience, etc. Move on,
suddenly that those assets become secondary. Has glacso managed to
(11:58):
change that buyer? Is it become? Are your systems in
such a way that once and as it comes in,
everyone has bought in and there is no champion left. Yeah?
Speaker 2 (12:07):
So well, and inevitably, you know, there are always champions
for individual assets when one finds them in the first place,
when you internalize them, and then they have formal project
leaderships leadership I expect those individuals to be, if you like,
the champions for their programs. But let me talk a
little bit through the process that we follow because in
(12:30):
each individual case, as I mentioned, we use the same
frame for judging internal versus external assets, and within that
there is an independently governed approach which would look at
the epidemiology, the market potential. We would access what we
call a probability of technical and regulatory success, and that
(12:50):
goes through individual expert teams and is then governed through
the same portfolio governance mechanism as I have for internal programs.
So to the greatest extent possible, the characterization of an
individual BD asset and the decision to prosecute it is
very much enshrined in that more quantitative approach. The championship
(13:15):
really comes in with regards to the energy and imagination,
if you like, we come with the design of plans
to accelerate and augment the value of individual assets.
Speaker 1 (13:26):
Great. So if I just have to touch on this,
it's an area where there's lots of discussion. I've been
doing some digging into the science space in China and
you can see there are there by any measure and
Nature does quite a good assessment of percentage of top
one percent citations, etc. And you can see that there
(13:49):
is a hockey stick effect going on there. But so
there's this discussion that there's been lots of licensing deals
from zero in twenty nineteen, let's say that's correct to
a particularly high number I have we're just finish finalizing
our assessment of the details, but several tons of deals
and you've done some yourselves, particularly ADC areas or t
(14:10):
sell engagers, et cetera. And so there's two things I
wanted to ask you. Number one is do you what
do you think has changed in your interactions with Chinese
biotaics that has enabled this surge. It's as simple as
the investment in science. And then do you see a
risk that some way politics will get in the way
(14:32):
of this, just as it's such a situation such as
the Biosecures Act.
Speaker 2 (14:37):
Yeah, And like Sam, I would say, probably the first
thing to say about m BD is obviously we have
a team that is global and we look to identify
innovation that is right for GSK wherever in the world
the science with the right characteristics emerges, and it would
be right to say that we're currently seeing an increase
(14:59):
in in volume of innovative science coming out of China.
As you say, we have a number of partnerships ourselves
and look in the context of managing the geopolitical situation,
Obviously we're very alert to whatever legislation is developing, and
we take account of that when we're making decisions about
(15:21):
individual assets, particularly with regards to the need to consider
ultimately the global manufacturing footprint that we have for one
asset versus another. So I would say this is just
simply part of the emerging and changing landscape of global science,
and we have the right mechanisms in place to spot
(15:42):
new areas of innovation and also to manage across the
legislative framework.
Speaker 1 (15:47):
Okay, so let me come to a particular asset, a
resurgent asset within the GSK portfolio, which is blend rep.
And the reason I mentioned this under our strategy and
broad discuss is clearly the first time round there was
something that was missed in the R and D program
in terms of the way that maybe do seeing whatever
(16:10):
the attitude was that ended up getting us to a
point where the side effect issues and then the trial
as a single agent didn't work out, et cetera. And
then of course now we have dreams seven and dream
made with very strong overall survival benefit. Just for the record,
we have a very deep analysis of multiple maloma and
we are big fans of blend rep. We think that
(16:32):
is a particularly well placed asset in the armamentarium of
treating a multiple myloma. So what do you think what
the lessons that any R and D organization And of
course in your case, yours can learn from the process
which blend Rep went through.
Speaker 2 (16:47):
Yeah, and look, let me just sort of reflect on
a few of the points. So you've made sam and
for the benefit of your listeners. Blend initially had conditional
approval in late line my loma, and in conditional approval
cases one has to have a confirmatory study, and that
(17:09):
for blen Rep was Dream three and that was a
monotherapy blend Rep versus comdex standard of care combination therapy,
and it was in late line. And when we look
at that, I think clearly, although blen rep has interestingly
a number of mechanisms of action, the decision to compare
(17:30):
monotherapy with standard of care is something that was on
behind the ultimate Dream three failure. And also that we
were looking for superiority versus standard of care. So I
think we just sort of shot for the stars in
too great an extent. In the Dream three study, of course,
we had Dream seven and Dreammate going on as well
(17:51):
whilst that particular study was being conducted, and Dream seven
and Dreammate were in the second line setting, so less patients,
and they were blend Rep as part of standard of care.
So if you like a more realistic comparison, and as
you mentioned, you know what we got from that, fortunately
(18:13):
was a fantastic result in terms of efficacy with blend
rep one, tripling the progression free survival for patients in
second line and reducing the risk of death even so
by forty two percent. And it's really unusual in MYA
loma to have both pfs and OS and this is
statistically significant OS. So we reported at the end of
(18:36):
last year. What we were also able to do through
the Dream seven Dreammates and indeed other clinical programs was
gain a better understanding of the ocular side effects of
the medicine. Dream three was the first time that those
side effects became apparent, and of course, over time what
one us learns more effectively how to deal with them.
(18:57):
And what we know now about the the blur vision
is that it's some reversible, transient and manageable and in fact,
in the vast majority of patients on studies, it's possible
to reverse the effect by changing the treatment schedule. And
if I just refer back to Dream seven, what we
(19:19):
know there is that more than seventy percent of the
patients had some form of schedule modification to manage ocular
side effects, and we still had those fantastic efficacy results
that I discussed earlier. So I think it's really just
about learning about a new medicine. You'll appreciate that learning
(19:40):
to manage the side effects of new cancer measure medicines
is not something new. You know, there are a number
of cancer medicines with significant side effects interstitial lung disease,
cardiovascular effects. So the cancer physician community are very used
to managing side effects. As they say in this particular instance,
(20:02):
that they're reversible and transient and very manageable.
Speaker 1 (20:07):
Right, great, yep. I have to say I agree with
most of that, given all the conversations we have with
physicians and various surveys we've done. So let's just to
wrap up on the broader strategy. What is Is this
something you can point to that's different about GSK's R
and D focus or R and D strategy to peers.
(20:29):
I mean, there are lots of peers, so it's unlikely
that there is any significant difference, But is this something
you wanted to point out before we moved on to
other topics.
Speaker 2 (20:39):
Yeah, and I won't repeat the focus across individual therapeutic areas,
but I think for us it's very much. We talk
about the approach of science times technology times culture, and
it really is the embedding of technologies, particularly in our
early stage portfolio, but also bling development that I think
(21:02):
sets us apart, and a culture as we touched on
earlier as well, that welcomes innovation both internally and externally.
And the point that I made with regards to external
partnerships and later stage assets also applies to our earlier
stage portfolio. You've no doubt quollowed that we had a
range of early stage collaboration deals that were announced towards
(21:25):
the end of last year as well, things like Relation
Therapeutics OPRAH or the Salius and we can talk about
each of those, but at their heart, these are biotech
organizations who have the same approach to technologies we do
at GSK. That is, the underpinnings of target and patient
choice coming from human genetics, and then the reinforcement of
(21:49):
those observations in what we might call the phenotypic features
of disease. I think those two things together give a
unique insight into choosing the right target and matching them
to the right patients at the right stage of disease.
Speaker 1 (22:03):
So would you say last question on the strategy R
and D et cetera. Would you say that that is
where the market underestimates GSK's R and D approach or capabilities,
not not specific assets, because we can talk about those
consensus versus company and I think at the JP Morgan
conference and at the Early Stage session that he had
(22:24):
just before Christmas, you highlighted the whole list of those.
But in terms of the broad potential for GSK's R
and D, is that where most people are underestimating or
is there another area you would like or areas that
you would like to highlight?
Speaker 2 (22:37):
No, would I would just say the look in general
consensus value is really focused on the parts of the
portfolio that the market can accurately model, and we might
talk about differences across particular assets. We've covered one indeed
with blend Rep, where I think the market is likely
(22:58):
waiting to see the nature of the lake that we
get with regards to REMS. And that's understandable given the
history of the asset that you've just described. They don't
tend to assign a great deal of value to early
stage technology, and it's why programs like for me, the
one that we started our conversation with the arrow eight
(23:20):
THESK ninety ninety molecule, that should we be successful with
a phase two proof of concept there that I can
point to something that not only brings value MASH and
alcoholic liver disease in particular has significant epidemiology. I think
it's the largest cause of liver transplantation worldwide, and if
(23:42):
I remember correctly, something like twenty six million individuals with
some form of alcoholic liver disease for example. So if
we can show with programs like that that we have
the marriage of scientific opportunity through genetics and function genomics
and the application of new platform technologies like all nucleotides,
(24:06):
and to marry that to opportunity, then I would hope
with increasing examples like that the market we'll see that
our approach is one that is indeed reducing risk and
enabling us to identify significant areas of epidemiology scientific opportunity.
Speaker 1 (24:23):
And from that value, I think this is a perfect
bridge to talking about technology and AI applications, et cetera.
So just let's ask you one hopefully very straight simple question.
There's nothing simple in this world. I guess, where do
you think we are in our ability to extract value
from AI and mL under curve. Would you say we're
(24:46):
in the first innings, whichever metaphor you want to use.
Speaker 2 (24:50):
Look, I'd say we're very much just at the bottom
of that S curve. And you know, we talk about
the S curve for technology and in the past we've
said it can be ten or twenty years before you
reach the top. I think what's going to be really
different about AI is that timeframe is going to be
massively compressed. Ultimately, though, I think what is really important
(25:11):
is to understand that the AI is really only as
good as the data that informs it, and that certainly
underpins our strategy. But I would say they're already and
we're seeing it internally as well. Areas where AI is
having a significant impact on the efficiency of what we do.
The design of small molecules, for example, is an area
(25:33):
in point, which is one that has been pursued both
internally in GSK and externally, and again that's a reflection
of where people can find high quality data. Protein structure
designed to go back to alpha fold, the PDB was
a source of information for the derivation of alpha fold.
(25:55):
The use of generative AI for the construction of regulatory
documents that I think is is already proving to be
an area of great advantage and efficiency. And again it's
one that we are also pursuing.
Speaker 1 (26:12):
Can I just stop you there, because you just mentioned
quite a few things. Let's take them one at a time,
and I'd love your perspective in this. Let's start with
the one that's probably the and you shoot me down
if you want, because I might be just talking rubbish here.
But preparation of regulatory documents is that one of the
low hanging fruits with the use of generative AI.
Speaker 2 (26:34):
Yes, I think it is. I mean clearly, in anything
you do with the regulator, one has to approach that
in a careful fashion and in partnership. But these documents,
wherever you can see an increase in quality and being
able to remove human intervention helps to improve the quality
(26:54):
of those documents. So that is something that works very well.
Of course, what it requires is the flow of information
into the document in the first instance, as well as
you know, the authoring of the document, if you like,
is merely the last stage in a complicated process. But
it is proving to be an effective approach. It obviously
(27:16):
requires human intervention at the end, because the thing still
may make errors, but it is improving efficiency enormously. I'm
more excited in clinical trials setting about some of the
other opportunities that AI.
Speaker 1 (27:32):
Right, right, let's let's let's come back to that in
a second. But I wanted to to see if you
would put a number on it. Tomorrow, say dream nine
top line is announced with the help of AI, how
much faster can you get the data to regulators than
you would have done without it?
Speaker 2 (27:51):
Yeah, and I don't think I've shared that data externally.
Sounds so what I'm not?
Speaker 1 (27:56):
But what's your What would you say if I put
thirteen heads of R and D the rough number would
be not necessarily GSK.
Speaker 2 (28:04):
I'd say, you know, they'll all substantially say that this
could come down to a matter of perhaps weeks. What
I would say for the final preparation of the documents,
what I would say internally is that when we look
again in our performance with comparators, then we're very well placed.
(28:25):
In fact, I think in the last CMR analysis we
were sort of second in the large Farmer group in
terms of that stage of the regulatory process. But as
I say, I think there are much much more exciting opportunities.
Speaker 1 (28:38):
There, right, So let's step back from that, because that's
the Actually, there's two sides we can go. We can
go forward and look at the patient identification and making
sure that the drugs get to the right physicians and
patients best educated fast as possible. So that's the launch
side or the step before it, whichever one you to
(29:00):
talk about next, which is the actual conduct, patient recruitment,
and the conduct of the trials.
Speaker 2 (29:06):
I'll take it in pieces, because let's let's do that next.
Patient conduct and recruitment of trials, the performance of individual sites.
This is an area where I think it is going
to be it already is in fact possible to improve
clinical trial recruitment. Behind that as well, there's an increasing
(29:26):
ability to examine a wealth of data that we might
refer to as secondary endpoint that determine, if you like,
the potential trajectory or risk factors for individuals, not necessarily
always encountered in the in the clinical trial design up
(29:47):
to now, but an ability you know what, one of
the things we often find in our in our studies
is we have a statistical hierarchy for analysis that requires
that you have points listed in such a way that
you have to succeed in each of them in order
to submit the subsequent one into a label. I think
(30:08):
AI is going to be very helpful in ensuring that
we understand how individual if you like, heterogeneity across broader
patient populations impact those types of outcomes. It's also interesting
to note that for a lot of our baseline characteristics,
people are judged at the value at which they are
entered into the study, and yet those characteristics themselves are
(30:31):
also not stable, and so you're looking at a trajectory,
and AI is doing a much better job increasingly beginning
to be able to project those trajectories. So there are
a host of different ways in which one might imagine
not only improving the recruitment through the performance of individual sites,
if you like, but also improving the characteristics the power
(30:53):
of your studies by better understanding the likely behavior of
patients and control arms. There's a lot more to be
done on this SAM, but I think you can already
see some of the opportunities that are emerging.
Speaker 1 (31:07):
Can AI replace the DSMBs in a live fashion if
AI is cognizant of the data that's coming out on
a daily, hourly basis, from the trials. Does that constitute
a problem for the regulator if no one has got
a view on that data?
Speaker 2 (31:24):
Well, no, I would say. In all of this, what
AI is doing is positioning those decision making bodies in
a better place to make higher quality decisions, more effectively
and indeed more quickly. So I don't see it as
replacing any of the independent decision making bodies. It simply creates,
I think, a higher quality proposition for them.
Speaker 1 (31:46):
And what are the issues? To use a phrase that
I think a lot of people hate to use, but
it was a good one, is the unknown unknowns? Is
AI better at connecting the dots in areas where humans
perhaps have bias and therefore don't look for it? So,
you know, the example being we always look for our
keys under the light. So do you think that could
help in terms of identifying issues or opportunities earlier as
(32:10):
the data flows.
Speaker 2 (32:11):
Yeah, well, look, I mean, let me step back from
the clinical trial environment and talking about this, because I
let's go first of all to where I think in
the early stages of the R and D process, AIML
again is showing its value already, and that is, to
your point, the ability to connect across vast data sets
(32:35):
near one level. Let's consider, for example, the use of images,
which are becoming increasingly prevalent as endpoints in early stage research.
I've been amazed by how good image analysis is at
getting at underlying elements of pharmacology. Now you need an
ability to qualify the components of the image that the
(33:00):
AIML is using to highlight changes. You can we call
them latent space variables, because an image contains a huge
amount of data, and so if you like separating signal
to noise, requires that you have the tools in order
to do that. But once you've been able to establish,
(33:21):
if you like, an abstraction of the image, which is
a faithful representation of the pharmacology that you're looking for,
then it is tremendously effective at finding features and indeed
potentially at identifying aspects of farm molecule performance that you
may have not initially spotted using a more specific endpoint.
(33:44):
Of course, in all of those cases you have the
reassurance that whatever AIML suggests, you can qualify your answer
by running an experiment. And that's why I went to
that area of research first. In the clinical arena, of course,
one has to be very very much more aware of
the potential risks associated with the use of an AI approach,
(34:08):
and of course there is emerging regulation there and we
remain very aligned to that approach and ensuring patient safety.
Speaker 1 (34:17):
So what do you think would be the poster child
of the first asset that is moving forward within GSK
that has leveraged all of these areas did genomics. I've
got a feeling you might say it's the alcoholic liver
disease or asset.
Speaker 2 (34:32):
But well, ultimately, if we get what I would call
clinical proof of concept, we have early results showing that
that molecule reduces the transcript for the targeting question and
that we reduce ALT, which is, as I'm sure you're aware,
a marker of liver injury. Perhaps we might come back
and talk about it next year and I see what
(34:53):
it does deliver fibrosis. But an early example in the
portfolio where we view AIML would be our treatment for
chronic hepatitis B for example. This is an interesting area
because in order to achieve what's called functional cure, one
needs to reduce the level of viral DNA and to
(35:18):
reduce a surface mark or tent called the surface antigen
to a level which is below quantification. Now for our
particular molecule, this is another Oligan nucleotide as it happens
at the reversant, it's the only molecule that is able
to do that in patient to a significant effect in patients.
(35:40):
We showed this in phase two studies where we had
a greater than fifteen percent reduction of both of those
features below quantifiable levels. Our models, though, which are based
on not only those factors but other liver measurements, show
that you have to get much lower than than the
quantum fiable limits to see a functional effect. And what
(36:03):
we were also interested in was the interplay between immune
system changes and the if you like, more traditional virology.
And what we've been able to do there is to
use a very is to use an approach with AIML
and imaging. On top of those are the quantitative measures
and that we feel is going to give us an
opportunity to perhaps help up to fifteen percent more patients
(36:25):
than just using the traditional approaches. So, and you know,
I could mention other areas that we're developing, but that's
the general theme that you will see come through that
we're able to better qualify patient response by a combination
of both image related data and complex molecular markers.
Speaker 1 (36:46):
Great. So, just to summarize all this, if I were
to look to ask you where the biggest impact will be,
Let's say you have the percentage that you wanted to
put on it, and it has to add up to
one hundred percent improved probablity of success, faster time to market,
and faster speed to reach peak.
Speaker 2 (37:04):
Yeah, man, let me sort of. And I think, of
course a lot of that depends on what point in
time are you going to judge it. I think, in
the first instance, what we'll see the efficiency gains, so
faster time from study to file, for example, you also
see efficiency gains in the discovery of the initial molecular
(37:28):
entities in the first place, small molecules or proteins. GENAI
works there, and it's not unusual to expect that you
might see our perhaps and I'm taking a very ball
part for you here, rather than our own internal data
that being reduced by as much as some fifty percent,
if not more. I think in time, though, what we'll
(37:49):
see is an improvement in our ability to match molecules
and targets to patients. And that's really what the important
goal is. You're a preciate that our Phase two success
rates across the industry are still relatively low. I mean
typically it's a one in twenty success. We talk about
(38:10):
the fact that genetics might improve that by twofold. Interestingly,
some of the work that we're doing looking at single
cell phenotyping, which is where this ability to use AIML
to layer different levels of data come in, they also
suggest that you might see perhaps a two or threefold
improvement in phase one. So I think ultimately the big improvements,
(38:33):
and name these may be in the order of a
fivefold increase in success will come in phase two studies,
where we typically at the moment test whether or not
our initial hypothesis was correct or not great.
Speaker 1 (38:47):
So I think I've answered every element that I just highlighted.
So is the implementation of AI and mL going to
be absorbable save by the industry as a whole within
current R and D budgets or will we have to
have a rise before we see the benefits?
Speaker 2 (39:05):
Well, at least the approach I can I can speak
to for GSKS were very much absorbing this within the
context of our existing R and D budget, And you know,
the change requires as any sort of change program does that.
I've always found that it's not only a matter of
adopting the new, but it's a matter of stopping the old,
(39:26):
and that that gives you an opportunity then to realize
these these changes. The big the big change for me,
that is the generation of more relevant, if you like,
human subject based data, and as I mentioned earlier, you
can see that in some of the collaborations that we've
put in place towards the end of last year, as
well as a more longer lived investments we've made in
(39:49):
human genetics. So all of that I think allows us
to make more effective decisions early on through the use
of highly autom mated systems, so we get to analyze,
if you like, more potential programs more rapidly, and then
in terms of the shape of the funnel, to be
(40:10):
able to pick ones that are going to succeed with
higher likely subsequent probabilities. Means that when you get into
the clinic, then one is able to run perhaps a
slightly smaller early stage clinical profile portfolio sorry with greater confidence.
That's where ultimately I think the change in the approach
(40:31):
that we will take will be most reflected in the
overall characteristics of R and D performance.
Speaker 1 (40:39):
So totally I'm conscious of the time. So I want
to help us wrap up here. Of course, I'll give
you the last word afterwards if you want to say
something that I haven't asked. But in the world of drugs,
in the years that I've been following it and analyzing it,
we've had darling areas or areas where they've been they've
become multi blockbus areas. So you start it off with
(41:00):
I think an area that GSK was involved in themselves,
the antacit world time. Then of course we moved on
to the to the lipid lower engagents, and then came
the era of anti tnfs and equivalents, and then obesity
now and is Glaxo do you believe invested in one
of the therapeutic areas that will become the next big
(41:22):
whatever you want to call it multi blockbuster and which
one is it?
Speaker 2 (41:26):
Yes? So I'm bound to say yes, aren't i? Sam?
And look, I mean the only one I would add is,
of course the PD one io in cancer.
Speaker 1 (41:35):
Of course, yes, sorry I missed that one.
Speaker 2 (41:37):
The characteristics of those areas as they tend to be
the culmination of high medical need or epidemiology, a new
understanding of biology, and often a new medality. Right you
You you self described the journey from small molecules into
into monoclonals in your in your summary of progress. I
(41:59):
would hope now that with some of the tools that
we've just described, we'll be better able to forecast those
areas than we have done in the past, because the
one thing I would say about that history is that
it's not usually the same organization that finds success. The
industry as a whole has succeeded, but individual organizations have
(42:20):
not repeated that. I would point to perhaps two broad
areas that where laying the foundations. I would say in
beginning to experiment in the first one, we've already covered,
and that's some tissue fibrosis following inflammation. I think somewhere
in the region of almost forty percent of deaths teaches
(42:40):
some form of fibrotic organ dysfunction or failure. And you
could almost within that include neurodegeneration. Where we're beginning, because
I think the mechanistic pathway to newer degenerative diseases is similar.
It's inflammatory and in nature and ultimately involves the consequences
(43:06):
then of that on loss of brain material and function.
We're beginning to invest in that area. We have collaborations
with Elector looking at progranulin modification. This points to a
very common area of biology and neurodegeneration.
Speaker 1 (43:25):
Is that often talked about that particular relationship.
Speaker 2 (43:29):
No, well, this is lyce'somal dysfunction, which I think does
is definitely mechanistic of mechanistic importance. We have obviously and
we are hopefully be able to say more about this soon.
Ongoing collaborations with health authorities looking to design not typical
(43:50):
perspective face three studies, but nevertheless perspective studies to evaluate
whether or not shingles vaccination protects from neuregenerator.
Speaker 1 (43:58):
I took one as soon as I said, read paper.
Speaker 2 (44:02):
And so those sorts of areas you'll see us continue
to experiment in. And as I mentioned earlier, the general
realization that the journey from autoimmune disease, which of course
is a cause of inflammation in and of itself, into
examining fibrotics sequally is something that our portfolio will begin
to take a closer look at. You can see the
(44:25):
beginnings of that in our work in lung disease in
liver disease, and we have some earlier programs in in
other end organs. We haven't disclosed them externally, so I'll
leave it at that. But there's a general theme here
of emerging new biology behind fibrosis and a new approach
to the newer degeneration.
Speaker 1 (44:47):
Did you I want to let you finish off on
any topic that you wanted to wrap up on. If not,
I have one last question and we can. It's a
very simple yes.
Speaker 2 (44:55):
No, And I think you've done a good job of
a journey through.
Speaker 1 (44:57):
Okay, twenty twenty five. Will we have an opportunity to
see IRA equalizes between small molecules and biologics in terms
of negotiation time?
Speaker 2 (45:06):
Oh? Look, I mean I'd say that is a matter
of discussions with the new governance in the US, and
all I would say there is Look, we've worked with
the Trump government in the past, we work with them again.
US continues to be an important area of innovation for US.
We've just put eight hundred million pack dollars into a
(45:27):
new film finish capabilities. So I generally tend to spend
my time focused on the portfolio. I account for the
various legislative differences that appear and and honestly, I think
you know, we still find a path to exciting programs
within whatever context that we have there. And probably the
(45:49):
thing I would say to finish is just on that
that I'm looking forward to another exciting year here. We
had thirteen successful Phase three studies last year, and that
of course ECHO is out now into five important launchers
heading our way, not the least of which, of course,
with a producer date in the middle of the area
is is blend rep in second line my loma. So
(46:12):
really looking forward to seeing how that works out. And
I've enjoyed our call today. Pleasure to spend time talking
to you.
Speaker 1 (46:20):
Thank you very much, Tony Wood, chief Scientific Officer at
GLEXO GSK. You have to get out of the habit
of saying that, thank you very much for spending the
time with US user US and
Speaker 2 (47:06):
USAGES
Host
Jonathan Palmer
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!