The Next Big Thing in Cancer Care

Dr. Lillian Siu:

Thank you, Peter. It is a great honor for me to be a part of this webinar. I will share my screen.

Good morning, good afternoon, and good evening, everyone.

I have the pleasure to speak at this webinar, and to be followed by two esteemed speakers, Dr. Lebwohl and Dr. Chua.

I will try and take this topic, Drugging the Undruggable, for the next 15 to 20 minutes. We've picked these topics because these will be the exact topics that will be presented at the Yokohama Breakthrough meeting.

Of course, the contents will be different, but we wanted to really whet your appetite now so that you will register for the meeting today after you've finished hearing the webinar.

I have the honor of being the chair of the program committee for the breakthrough meeting in August. I think this will be a very exciting meeting because the agenda has been well-planned for the last year or so.

The speakers in the meeting is going to be extremely-- the speakers are going to be chanextremely excellent selection to showcase.

Let me move my slide. These are my financial disclosures. As I mentioned, I have highlighted this topic specifically because it will be in the breakthrough meeting in August.

This particular session, Drugging the Undruggable, will be chaired by Dr. Pasi Janne from Dana-Farber as well as by Dr. Kohei Shitara, who is from GI Department of the National Cancer Center in Tokyo, Japan.

We have three topics in this particular session in Yokohama, the first one being in the Current Status and Future Directions in Drugging the Undruggables. We will also talk about targeting KRAS beyond P12C as well as G12C, with new and upcoming.

Then, lastly, we will talk about a Novel Molecular Glue for MYC-Amplified Cancers. I will attempt to really just highlight a bit of what I think will be discussed, but I think the actual meeting itself will be far more exciting than what I am going to show you today.

Please do come to the meeting in Japan.

Obviously, as a drug developer, we have been drugging the druggable for decades. It's been successful, but I think what would totally open a whole new era in the field of developmental therapeutics is really trying to drug the undruggable.

By the term "undruggable," the definition has been really a protein that is not pharmacologically capable of being targeted.

That, obviously, has been a challenge for us in terms of trying to find medicinal compounds that can actually do the impossible.

Why are they undruggable? Typically because the surface of these molecules are just too smooth. There are no pockets for the drugs to dock, and there are no pockets to allow even protein-protein interactions.

Because of that, it's very difficult to develop specific compounds, either to block the oncogenic proteins, or to activate the tumor suppressor to make the tumor suppressors work again.

Some of the examples of the undruggable targets I have highlighted here are, for example, the transcription factors such as MYC or the STATs, STAT3, for example, STAT5, many of the tumor suppressor genes that we know very well as well, that we have been trying to drug for the last many years, such as TP53.

Also, a new set of molecules are the deubiquitinating enzymes, or the DUBs, really are starting to emerge as becoming druggable from the undruggable category.

Many of these have now started to move towards that side of the pendulum to make them become therapeutic targets for us in the drug development world.

Now, I'll start off by talking about something that already has entered and got FDA approval.

The KRAS G12C inhibitors are familiar to many of you because they have been approved, at least using accelerated approval mechanism by the FDA.

The two listed here are Amgen's Sotorasib and Mirati's Adagrasib. These two have both been approved in non-small cell lung cancer in those who have the KRAS G12C variant in lung cancer, which is only about 14% of the patients with non-small cell lung cancer.

As you can see, based on the published data, the response rate of these compounds are about 35% to 40%.

The medium PFS are actually quite consistent across these two drugs, about seven months, and the median OS, about 12.5 months.

You will know, obviously, the whole story of Sotorasib has led to the FDA put out a guidance in terms of Project Optimus because the recommended dose of 960 milligrams was the highest dose that was tested in the phase one trial.

There is a post-approval commitment from Amgen to actually look at a lower dose in comparison to the approved dose to really understand whether we do need to give a dose as high as the top dose or whether we could actually lower it and still achieve the same efficacy.

You also will remember the data from CodeBreaK 200 which was just presented in ESMO this year, I should say last year, 2022, that in patients with pre-treated non-small cell lung cancer, Sotorasib compared to docetaxel had improvement in the progression-free survival, but not an improvement in median overall survival, even though the study was not powered for overall survival.

Therefore, obviously, that does give us some idea in terms of how best to develop this class of drugs.

Both of these drugs are being developed in combination with other compounds.

For example, trials that are ongoing with immune checkpoint inhibitors or with other targeted therapies such as a SHP2 inhibitor.

So far there has been challenges in terms of the therapeutic index when you combine these drugs with other agents instead of monotherapy.

These drugs are also being tested in monotherapy as well as in combination in colorectal cancer as well as pancreatic cancer. Because KRAS G12C is present in about 3% to 5% of colorectal cancer and in about 1% to 2% of pancreatic cancer.

The pancreatic cancer data have just been published in New England Journal of Medicine.

The colorectal cancer paper has published recently in Lancet.

There's no doubt that KRAS G12C remains a very exciting no longer undruggable molecule now druggable. There's probably many more to come in terms of this particular variant in terms of this target.

Now, there are obviously other undruggable targets of interest and I've just listed here. Once we have done drugging KRAS G12C, you can imagine other KRAS-specific mutations are becoming interesting targets for us to consider.

There are also lots of efforts to target Pan-KRAS, so not one specific variant, but a variety of variants along the KRAS gene family.

ven also in addition to that, the Pan-RAS so KRAS, HRAS and NRAS. If we start thinking about doing that, obviously the therapeutic index becomes the most important question because there is also wild-type RAS proteins in the body that potentially could be affected.

TP53, as I mentioned the most important tumor suppressor gene in our system is an important target if we can drug it and make it active, makes it very exciting.

Listed on the slide are all the other targets that are also being actively considered in the undruggable arena to see how we can make compounds to drug them.

I won't go through all of them tonight. Moving still a little bit in the KRAS world, I've mentioned already KRAS G12C. Obviously, KRAS G12D is actually the most common KRAS G12 variant. Obviously, it is a really important target for drug developers.

Then all the other KRAS variants are also on the radar. As I only listed here in a recent article in Nature just published a few months ago, highlighting some of the drugs that are actively in development.

Sotorasib and Adagrasib data I have presented earlier, but you can see there are others that are also still targeting KRAS G12C, potentially making it more effective in terms of longer duration of response or longer overall survival hopefully.

Also more combinable with other agents, which have been the main challenges with KRAS G12C inhibitors and obviously in also overcoming resistance because these compounds don't seem to work for prolonged periods of time.

KRAS G12D is the next target on the G12 arena, but there are challenges with targeting G12D. I will mention a little bit earlier and I mentioned briefly about the Pan-KRAS initiatives as well.

KRAS G12D, even though it is, as you can see, 36% of pancreatic cancer, it's a really large population for pancreatic cancer, which I would argue is one of the hardest cancers to treat in solid tumors.

If we could target and drug KRAS G12D, that would open a lot of interesting opportunities for these tumors.

Even there as you can see, 12% of colorectal cancer and about 4% of non-small cell lung cancer.

Two years ago, I guess a year and a half ago, the Mirati group presented this at the ENA meeting and they presented their compound targeting KRAS G12D, which is MTRX 1133.

This was a non-covalent inhibitor because covalent modifiers targeting aspartic 12 is just not practical and there is really not effective inhibition of KRAS G12D with the KRAS G12C inhibitors.

The GTPA's activity is much lower for KRAS G12D compared to KRAS G12C. In other words, it's just a much more difficult target than KRAS G12C. Nevertheless, they did present some data on this non-covalent inhibitor.

You can see on the slide they presented data on various xenografts of different KRAS variants and also they can see various xenograft models with fairly impressive tumor growth inhibition.

The specific compound does spare wild-type KRAS and seems to be quite specific on KRAS G12D. However, we can see that in the last year and a half, the development of this compound has been challenging.

It has not entered phase one trials and likely there are still going to be a lot of challenges with targeting KRAS G12D with a non-covalent compound.

We'll see what happens to this or other KRAS G12D inhibitor candidates in the field.

This paper was published fairly recently by a group in China looking at a different way, an allosteric way to inhibit KRAS G12D by creating a salt bridge that can turn off the aspartic, Asp12 in the position and it hopefully will create the switch pocket to allow it to disrupt the interaction between KRAS and its downstream effector molecules.

Hopefully, that would be another new way to think of allosterically inhibiting this particular variant, which is really, really a challenging molecule to block.

I mentioned about the Pan-KRAS and listed some of the compounds that are currently in either preclinical or in phase one development and these are not the only compounds.

There are many others also in creation right now.

Many of them have presented in different meetings as I listed in the slide below, looking at xenograft models and cell line models in terms of inhibiting across the entire RAS family and sparing the wild-type KRAS, which is obviously important to reduce off-target side effects.

I suspect in the next year or so, we will see a lot more in terms of RAS inhibition.

I want to point out that we did invite the two guy group to come to the breakthrough meetings since they are in Japan, in our host country.

They have a very interesting collection of medium size molecules and I think they will be presenting some data related to RAS inhibition using this particular class of compounds during the meeting in Tokyo.

Let's move onto TP53, which is really one of the most important molecules in cancer because it is inactivated in more than 50% of cancers across all different cancer types. If there's a way we can switch TP53 back on, it would be an important breakthrough in oncology.

At ASCO in 2022, this is presented by Dr. Dumbrava in MD Anderson. Looking at a small molecule that selectively binds to P53 Y220C variant.

Again, a very selective specific variant and its goal really is to stabilize it, and twist it back into the wild-type conformation and therefore restoring its function. This is by a pharmaceutical company P&B Pharmaceuticals or therapeutics.

As you can see in the waterfall plot, there were different tumor types that were studied.

There were activities seen in small cell lung cancer, breast cancer, colorectal cancer, et cetera.

The toxicity profile is actually quite variable.

There is a variety of toxicity that felt to be overall still tolerable.

I think we're going to see more later-phase development of this compound in terms of its way to activate TP53 to make it work in these cancers again.

An interesting paper that was published in Cancer Discovery a few months ago by Kevan Shokat, who really is a big name in Drugging the Undruggable because of his work in KRAS G12C, published his paper on, again, P53 Y220C specifically cysteine, targeting covalent bonding or inhibition of the cysteine molecule in the 22C space, which I think really is an exciting compound that we should watch because I think this will enter patients once they identify the most ideal candidates.

I think it will be similar to the story of Sotorasib and Adagrasib, that we should watch this space.

The idea really is to again rescue the wild-type thermostability and gene activation of P53.

Instead of KRAS G12C where we're trying to block, P53 is a good gene, so we need to activate. We need to stabilize it, and reactivate it.

In the data that was published by Guiley and Shokat in Cancer Discovery, they showed that this particular candidate seems to be able to do that.

It's tested several candidates, but the one that seems to be the lead of all the ones that they were showing seems to have that particular property.

Again, very exciting story. I think we will see more of this in the literature in the next few years.

The last bit of my talk is to also focus a bit on the degraders.

There are so far two big classes. The PROTACs and the Molecular glues.

The PROTACs are bifunctional molecules, so they are connected through a linker and it binds to typically E3 ubiquitin ligase and the target protein and degrade it.

The molecular glues similarly, but not exactly the same, they're typically small molecules.

They typically bind to either the E3 ubiquitin ligase, or the target protein, and then allosterically induce a protein-protein interaction and causes the breakdown of the target protein.

There are several of these kinds of drugs in the clinic already and some of them are in active development.

We will showcase this particular class of drug in Yokohama. Just highlighting here some of the targets that are of interest to the world of degradation.

Obviously, we already have drugs that are degrading androgen receptors and estrogen receptors but listed here are some of the other perhaps harder or more difficult to break down molecules as well that are actively being pursued by the degraders and the glues.

At the Yokohama meeting, we have invited Monte Rosa which is one of the company developing glues to present their data on drugs specifically focusing on MYC amplified tumors.

We know that MYC is really relying on protein translation. It's dependent or addicted to protein translation.

By developing this compound that specifically interacts with the protein translation terminating factor, GSPT1 is going to hopefully lead to degradation of this particular protein and therefore, stopping a lot of the activity of translation that is driven by MYC.

This already has entered patients.

This is currently in phase one trials and is specifically targeting non-small cell lung cancer, small cell lung cancer, high-grade neuroendocrine carcinoma, some of the lymphomas as well as other L-MYC and N-MYC amplified cancers.

These are the tumor types that have very high MYC-driven and high-MYC amplification when you do molecular profiling.

We'll hear a bit about this from some of the worlds of the molecular glues as well as protein degraders in Yokohama.

I'm going to conclude. I think there are going to be many exciting topics that will be discussed, but also focusing on Drugging the Undruggable I think we can now change the word to potentially druggable instead of undruggable.

Truly, this is the field that leverages the combined expertise of chemists and biologists as well as clinicians.

I hope you all will come to ASCo Breakthrough to hear more from others in this field.

I think you will not be disappointed by that. Just again, one word for you to please register after today's talks. I will pass it back now to Peter.

Dr. Peter:

Thank you, Lillian.

That's really certainly whetted my appetite.

It's nice to hear things on a Zoom presentation.

I think we're all thirsting for actually meetings that take place in person where we can have really in-depth conversations and really have undivided attention.

It's certainly the presentations you highlighted showcase that really are several new classes of drugs, new strategies for drug development with real examples of drugs that are in the pipeline entering clinical trials or in the case of KRAS are already available with further marketing work to look at the best dose.

Our second presentation tonight is from Dr. David Lebwohl, who I remember from the days when we were fellows together.

It's great to see how David has really pioneered breakthroughs, especially at his time in Novartis, where he led the global team on the development of the first CAR T therapy.

David is now the Chief Medical Officer at Intellia Therapeutics based in Cambridge, Massachusetts. He's continuing his developmental work in CAR Ts. He's here tonight to share a lot of really very exciting laboratory preclinical data, and the strategy that is being used to make CAR T cells more effective and readily available.

With that, David, you can take over the screen and start your presentation.

Dr. David Lebwohl:

Very good. Thank you, Peter.

I thank the organizers. I have to share first. Let me get that.

We'll get a quick preview of the whole talk. Thank you, everybody. Again, thank you for including me in this talk.

I want to talk today about CAR T cells which you know have generally been autologous cells coming from the patient themselves, and then how we can engineer them and perhaps take cells from a donor, an unrelated donor to be used in patients.

Here first looking at Kymriah, this was the first CAR T product that was approved in the world.

I was part of this development. Very interesting because this did start as an academic program at UPenn from Carl June's group about a decade ago, but then was licensed into Novartis.

We were able to produce this at Novartis using industrial methods in 2014 and had an IND submission. That was wonderful about what we've all seen at Penn.

Then the product that Novartis was able to make.

We had a greater than 90% complete response rate in patients with childhood pediatric acute lymphoblastic leukemia.

This has really lightened up a large amount of work in the CAR T area, and a range of other types of hematologic cancers so far, both diffuse large B-cell lymphoma, and also with the new target DCMA growing after multiple myeloma.

You can see how rapidly this can go because of the excellent activity we're seeing after an IND submission in 2014.

It was really a phase one study that showed similar activities for the Penn work cut-off in 2016 and the approval in the US in 2017.

How do you make CAR T cells? It has to be made patient by patient if you use autologous T cells. 

The patient there's the leukapheresis that takes place, cells is collected from the patient, then they are reprogrammed.

This is often done at a far distance from where the patient is because the manufacturing does need to generally be centralized. That takes several weeks.

In that period, of course, you'd have to make sure the patient is supported for this period. Patient then undergoes leukodepletion in order to prepare for the infusion of T cells.

The T cells are infused and the patient is monitored for successful engraftment. For leukemia, the complete responses we're seeing within a month. It is a very rapid response.

On the safety side, it can be associated with too much activity on the part of T cells causing a cytokine release syndrome that you may be familiar with.

People have learned very well how to control this for almost all patients.

There are some downsides then to autologous T cells.

First, the need for viral vectors for patient by patient. This is very time-consuming and can be inefficient. There can be inconsistent apheresis coming from patients.

You can imagine a patient who has a large number of blasts or a patient who has very low white cell count. Sometimes you are not able to manufacture for individual patients.

Of course, this is a very difficult situation then for that patient. This was generally needed to be done at centers of excellence.

There really were, especially when we first started, only a handful of centers that were able to work with CAR T cells.

This has luckily become much more distributed because of the availability of approved agents. It still requires some specialized knowledge at each of the centers.

An issue I won't talk so much about today or any about today. Other issues with any CAR T cell therapy is that there is escape.

Like resistance to other drugs, there can be escaped by in the case of CAR Ts directed at leukemia, there's loss of CD19, and then a resistance develops.

I want to start to talk about how can we do better for the patients by having a product that is manufactured coming from a donor, and therefore potentially be off the shelf for that patient.

What we've been able to show at Intellia and other people, I should say, are working on this. This is not the only way forward for allogeneic cells, but we do think now this may be the best way forward.

We use an LMP-based cell engineering, and I'll talk a little bit more about what that would mean. With that, we get highly efficient editing.

This is CRISPR editing. We think we can get very good cell performance which I'll show you. This is scalable as a process.

Instead of doing this patient by patient, you potentially could make a drug for 50 or 100 patients with a single donor's leukapheresis.

Now, this can be very versatile, various cell types as you can imagine, T cells, NK cells, macrophages and I will focus on CAR T today.

Of course, the receptors may be CARs, but they might also be TCRs or other types of binders. Another direction this can go because you can edit with CRISPR, you can go beyond just adding a new receptor to the T cells.

You can also enhance the immune function of those T cells by various edits. Again, I won't address this so much today. They're not needed so far for hematologic tumors, for example.

As you can imagine as you get to treat solid tumors, you will need to enhance the activity of T cells because of the inhibitory environment that they find themselves in solid tumors.

Now, how did we differentiate our approach? Part of this is lipid nanoparticles.

Generally, people are trying to edit cells by doing electroporation. You can imagine this involves in a sense putting electric field across cells.

It can be very damaging and has been very damaging to cells, but then very much healthier using lipid nanoparticles.

We do sequential editing. In the case of CRISPR since you are cutting DNA if you just cut one site at a time this is much safer.

It avoids the risk of translocation. It avoids any type of rearrangement of the DNA. That's the approach we've taken.

You'll see that we do get very low translocations. This can be done easily with a cleavase or a base editor.

In terms of insertion, instead of having a lentivirus, which does insert randomly into the genome, you can precisely insert into the TCR locus and therefore know exactly where your CAR construct is going.

Now, there are three key concerns in making allogeneic cell therapies. The first is the ability of the T cells you're making to attack the host.

This is graft versus host disease. We do need to solve that in order to move forward with allogeneic therapies.

The second issue is that T cells can be rejected by the host because of the HLA differences. In this case, it will be important, as you'll see, to match the HLA.

Now, the biggest issue with the current programs that are out there, and you've probably heard about different allogeneic programs being used, is that they often eliminate all the HLA on the CAR T cell. 

When you take away HLA, they become targets for natural killer cells. I'll talk about how we avoid that rejection, and that rejection occurs very quickly when a cell is given without HLA.

First looking at graft versus host disease, again, this is due to endogenous T cell receptors that would be on the T cells taken from a donor.

They will recognize the mismatched HLA Class I and II in the patient's tissues. In order to avoid this, again, you have the power of CRISPR where you can target a single gene in this case knocking out the TCR receptor.

In addition, as a second step, you can donate the CAR T into that same site that you're cutting in the TCR locus.

So, you're very specifically getting the CAR into the native locus for TCR. This now really dissolves the issue of graft versus host. There's no detectable endogenous TCR that remains after this process.

Moving to the second issue of T cell-mediated rejection. In this case, there have been several approaches taken so far.

Some of them involve lymphodepletion to try to reduce the allo reaction.

Sometimes there's been partial matching done, but that only reduces the rejection.

It doesn't get rid of the rejection.

We feel we have solved this issue particularly because we are matching the cells.

In this case we first knockout HLA Class II as well as knocking out HLA-A.

Then in order to have a matched set of cells to the PA to the patient, we match the remaining Class I alleles, which is B and C.

I'll show you a little more detail on how that works in the following slides. I do want to move to NK-mediated reactions.

Again, this happens when there's low or no HLA class on a CAR T cell. In this case, because we retain HLA-B and C, we do not see natural killer cells rejection of these cells.

This is a natural inhibitory process to avoid NK destruction.

Now this gives an idea of how you match, and this is, I should say, a little complex, but I hope I can have you understand it.

Imagine a patient has a certain set of B and C alleles for HLA and they will only be compatible with cells that have a matching B and C. The way we do match this is we find a homozygous donor.

We reduce the complexity of the B and C in that patient. They'll have one type of B, one type of C.

Then as you see on the right we could obviously match a patient who's homozygous for the same B and C.

We could also match patients who might be heterozygous for those B and Cs.

There's some examples shown with the green check mark. The only patient obviously ineligible for this if they have a B and C that's com that's different from the donor that we've created.

Now, of course, you're going to need more than one type of donor for this because there is a lot of variety in HLA-B and C and this case I'm giving examples from if you're in Hong Kong or Malaysia, also Indian and Sri Lanka.

What you see in all populations is, by having just 10 donor types you are able to match up to 80% of the population.

This would address a large part of the population who might be able to be treated with off-the-shelf type of allogeneic CAR T cells.

Now, to go on to some experimental data showing what I've said is working in preclinical models.

In the first case for graft versus host disease, looking to the right, you can see that the Allo TCR T cells that we have don't induce any signal of T cell reaction from the host, and this is shown as a level blue line.

However, if you have unedited cells, you do develop graft versus host and attack growth of those T cells indicating their proliferation due to a host mismatch and you see a very big difference from the Allo cells that we can produce.

Moving on to T cell-mediated destruction. This is now the host destroying the CAR T cells.

Again, looking to the right, if you mismatch the first unedited donor T cells, you get a lot of proliferation of the host T cells that show that they're actually rejecting the CAR Ts.

While if you match the next two lines, either taking away all the HLA or matching by taking away A and type two and matching B and C, you get very little activity from the host.

This will translate into the persistence of the cells that are Allo matched.

Now, the third issue, I just said that you can avoid the destruction by Allo by taking away all of the HLA, but this does not take away, unfortunately, the NK destruction of those cells.

They recognize cells that do not have HLA.

You see, if you look at the graph, this is showing T cell lysis and destruction of the CAR T cells. If you either remove everything with beta-2m, this gets rid of all HLA.

They're rapidly destroyed by NK if you get rid of HLA and actually give back an HLA-E which has been thought to be a strategy that might help in this situation, even there, you get a great deal of destruction by NK cells.

If you look at the bottom, Intellia's Allo T cells. These are not touched at all by NK cells in this model.

I'll show that again in mice, in this case in vivo.

If you look on the left, these are again B2M knockout, so they've lost all of their HLA. In mock mouse, you don't have NK cells.

These cells survive on the upper curve, but when there are NK cells in those mice, you very rapidly within a day get a loss of greater than 90% of the cells going more than a log reduction.

On the right, you can see that with the Allo T cells both in the mock mice or an NK engrafted mice, you get survival of the cells.

The cell is perceiving the presence of the B and C and therefore there's no destruction by NK cells. These cells are very healthy.

I talked about the use of lipid nanoparticles.

Of course, this is very much like the COVID vaccine that is so important around the world.

It's putting in the lipid nanoparticles, whatever components you want to get into the cell including CRISPR delivered by mRNA, just like vaccine is delivered, you can also deliver the guide RNA that helps direct it to the right place in the genome.

The exciting part of this. If you look at the expansion of the cells, you can get greater than 120-fold expansion.

They are quite healthy despite the editing process. The viability is very high.

Looking further to read the bar graph, the viability is high, around 80%.

You're getting 80% of the cells with all of the edits that you've introduced and that is three different edits. It's quite efficient.

You get high CAR insertion rates. Rating 80% of the cells have the CAR and you get very high levels of expansion including keeping their memory phenotype which is important for healthy CAR T cells.

If you don't do this, you may get exhausted cells in other systems. This is to show that these cells work very well. 

hey work as well as the autologous CAR T cells both in vitro on the left and in vivo on the right. They look precisely like an autologous CAR T cell.

Just to summarize the strategy. First knocking out endogenous TCR, inserting in that locus CAR so you get very precise and nonrandom integration of the CAR.

You knockout Class II, you knockout A, and then you can match B and C in the patients.

This addresses all the issues that we think are affecting the current programs looking at allogeneic CAR T cells.

Moving towards the clinic, we do look very much forward to talking to you about our progress as we move forward.

You can see that you need a small number of donors to match with B and C.

It's not expected to require intense immune suppression, we would use a standard regimen of capecitabine and cytoxan that's used for CAR T cells.

It can be used for any type of receptor really, CAR T also TCRs which have been used more in solid tumor models. As I mentioned, you can also edit the cells as another step to enhance their immune activity.

Thank you and I'll be turning this to Peter.

Dr. Peter:

Thank you, David. I'm looking forward to the panel session.

You've tantalized us with the prospect that this strategy may enter into clinical trials sometime later this year. We certainly would hope so. We'll look forward to hearing more about that.

For our final speaker today, we have Dr. Melvin Chua who is chair-elect of the program committee. He will be looking forward to our next meeting after 2023 and will be the program chair.

That is also chair of the department of head and neck and thoracic oncology at the National Cancer Center in Singapore, as well as leading a laboratory that has interests in the intersection of data science and genomics.

With that introduction, Melvin, I will turn it over to you.

Dr. Melvin Chua:

Thank you, Peter. Thank you.

Let me just share my screen.

Good morning from Singapore and good evening wherever you are.

It's a great pleasure for me to be able to share with you all some of the things that we have lined up and to speak with such esteemed speakers.

With that, I'll move straight away into the content of my talk today.

I'm going to speak about the two separate sessions we have lined up at the ASCo Breakthrough meeting.

Hopefully, with all that, I'll be able to convince everyone that's signing in today and also advocating to your colleagues to join us in Japan this year. Those are my disclosures.

The first session I will touch on focuses on multi-omics technology.

This session is going to be chaired by Professor Takayuki Yoshino from Japan and also our dear colleague, Professor Daniel Tan, who is also a drug development person practicing at the National Cancer Center Singapore.

To set ourselves a little bit different from some of the [unintelligible 00:49:05], NGS is not exactly new technology.

We've decided to focus on some of the more novel areas in this space ranging from liquid biopsies where we intend to focus on some of the pivots into new technology as well as potential clinical applications that are seen in some of the exciting data coming up in colorectal cancer, for example, and also in other tumor types.

Now, in line with the whole theme of this meeting, as being a showcase of innovative technology, we are also going to talk about some of the new attacks on single-cell sequencing, as well as digital spatial profiling that will provide some insights to the heterogeneity of tumor and in essence, the thoughts that we have lined up aims to help the audience envision how this tech can actually be used, not just in the setting, not just for translational research, but also in terms of biomarker discovery from clinical trials patients to being used in the clinic.

I think that's segues very nicely to one of the other sessions that we have, one of the talks that we have for this session that aims to discuss how we could actually bridge this technology into the clinic to give you some practical sense.

These whole talks is timely because for most of us in this field regards to industry partners to the practicing clinician is that we've seen an embrace of multidimensional data and oncology.

This is a very nice summary illustration, very nice review by Peng Jiang authors whereby they showcase how the different layers of data where we essentially move from electronic health records to now because of the digitization of health records now transitioning from historical notes, physical taking notes to digitized records.

This then allows us to think about bringing in things that could in fact also be digitized, like imaging, and then the whole sort of segue into the molecular realm whereby we talk not just about the genomics but also the gene expression, the proteomics, and more into the circulating molecular profile domain.

When we have all this data that comes together, I think it is exciting as it helps to accelerate some of the work that we do in terms of research, in terms of bringing new drugs to the clinic.

I think the whole idea is that this also helps the field that needs to think about how do we actually better integrate all this data and do we actually need all this data for treating a patient in a clinic?

Of course, it's great that we have all this data available, but I think with the cross intersection of all this data I think, and with the consciousness of cross effectiveness in today's clinical practice, I think it's important then think about what information might be best use for treating the patient.

Of course, some of the motivations behind the move into, look at biopsies for example, speaks to some of the limitations with current technology in terms of what we face in a clinic often as you said on discussions at tumor boards and all that.

We often get faced with the practical limitation that the patient may not have tissues available for profiling, and I think conversely on the flip side, when we talk in terms of molecular NGS data, a lot of times we get information that is aggregated or derived from tumor samples, which is essentially homogenizing mesh together, where we may not be able to appreciate the intratumoral heterogeneity in terms of drug actionability.

Ultimately, I think a lot of the how much of this information actually translates to improvement in patient care remains a perennial issue that's faced by most of us.

These are some of the popular solutions and hence why we thought this was timely to then bring up these topics in our session at ASCo Breakthrough whereby the film may think that liquid biopsy could be a nice solution if it's able to corroborate with information in a tumor in terms of actually deriving actionable tumor information, molecular information.

Now in terms of addressing bulk-level data, I think I've mentioned the single-cell profiling and more recently digital spatial profiling could be technology that could help address this issue.

Of course, to the practicing clinician, one often posit if it is this technology is primetime for use in the clinic or even in the design of translational studies from clinical trials.

Starting with liquid biopsy, I think the utilization of liquid biopsy spans across the entire oncology spectrum. In this nice illustration published by Dr. Michail and authors at Nature Reviews Clinical Oncology, they showcase that there are different dimensions of liquid biopsies ranging from circling tumor cells now moving towards the ctDNA and more recently circulating RNA which may inform on the transcriptome the circulating transcriptome which may better guide us in terms of therapy selection.

Nonetheless, I think regardless of which platform we go with, the idea is that if you have a very sensitive and potentially accurate tool, then it could be used across a different clinical setting in terms for screening for early cancer detection.

I think that's where a lot of the focus has been because with the notion that if you detect a cancer early, potentially that might help subjugate your course of treatment and also better outcomes.

Now in the other two scenarios where we often, more we see a wider use of application of liquid biopsies in today's clinical setting is in the surveillance of recurrence as well as for treatment selection.

I think in one of the tumor types that I treat, nasopharynx cancer, we've been very fortunate to have a circulating tumor DNA biomarker that not just allows us to prognosticate and monitor treatment response, but in the surveillance setting, we now have this ultrasensitive test that could actually be done regularly and actually help to interface with our conventional staging investigations like PET-CT and CT which one would imagine that you can't do a scan every three to six months, and so, therefore, this allows such a tool to be nicely inserted into the clinical practice domain.

I think in terms of drug selection or treatment adaptation, which is a very novel concept in terms of how we think about treating patients these days, could we actually intervene early?

I think this is where we hope that liquid biopsies would be able to offer this information because it seems almost unreasonable or illogical to be able to do sequential biopsies of a patient's putting them through multiple invasive procedures.

This is a very nice paper I'd like to showcase and certainly, it'll be discussed at this year's ASCo Breakthrough, whereby we now see tangible application of this technology in the clinic. 

his is a very nice New England Journal of Medicine paper whereby the investigators from Australia as well as in the US attempted to answer the question of using circulating tumor DNA to actually guide adjuvant therapy in a very niche subset of patients with stage II colon cancer.

And the idea is as such, so this is a group of patients whereby right now, it's uncertain if they'll benefit from adjuvant systemic therapy, so the investigators try designer study whereby in a 2:1 randomization, a group of patients will be treated as per physician discretion based on commercial clinical pathology parameters to then decide whether you should receive adjuvant chemotherapy or not.

In the ctDNA-guided group, the investigators would then have ctDNA information, and then if it's a positive ctDNA result, despite the clinical pathological parameters, patients would then be referred for adjuvant systemic therapy versus for those who have a negative test, then they'll be observed.

These were very nice results whereby investigators showed that with this tool alone, just by implementing this tool in the treatment decision workflow, they essentially were able to achieve similar two-year recurrence rates between the CT-guided management decision as well as standard management.

At least, it's showing that using a liquid biopsy to guide treatment decision, it is no worse than what we currently do conventionally in a clinic.

I think what is also impressive, if you look at the bottom panel in terms of saving or rationalizing adjuvant systemic therapy, in the standard management group, it will seem that a much higher proportion of patients receive adjuvant systemic therapy.

Potentially although the trial is not really set up to address the question, if you use ctDNA, you could reduce in a singular patients the use of adjuvant systemic therapy by about half.

In the population setting where colorectal cancer is a fairly common cancer that affects everyone around the world, I think we could imagine that such a tool that was cost-effective and implementable could have far-reaching cost-effective benefits.

Now this to fueling night excitement.

This is a snapshot of all the industry partners and commercial activity, and the research that's happening in this liquid biopsy space that spends across all the different clinical scenarios that I mentioned earlier, renewing from early detection.

You see a lot of the efforts really are focused on surveillance as well as using the same tools potentially for treatment selection.

I think needless to say, with the further amount of activity is happening in this space, this is a huge market opportunity for both academic as well as industry partners alike.

Now, pivoting to the other new technology that I briefly spoke about, just to elaborate a little bit on this single-cell profiling.

I think single-cell profiling essentially allows us to look at the tumor dissociated in the single cell and allow us to actually understand the profile of the cell, both at a genome as well as a transgene expression level at the single-cell level.

Potentially what it does is that, apart from the excitement of actually knowing the single cell, the molecular data of the single cell, I think the fuel has since moved on quite rapidly in terms of how do we actually make sense of this data.

Typically today when we profile this tumor at the single cell level, the true bioinformatics pipelines, the researchers are able to then by virtue of the similarity of the characteristics of each cell, we then able to generate a two-dimensional map like that using the multi-dimensional data that we have at a cell level to then be able to derive clusters of cell population that are most similar to each other.

That allows us to at least be able to understand some of the classes of population, we then design potentially therapies for.

Now, the technology of digital spatial profiling takes this a step further, whereby you're now able to visualize this sub-population, apart from just dissociating them, and then profiling them using conventional single-cell techniques where one of the key issue has been cell attrition.

Then with digital spatial profiling, this addresses the issue of actually sample attrition.

Potentially you get a high yield of data, but the idea is still the same whereby you're not able to image cells at the single cell level and to be able to stain them with multiple immunostains. This is a nice example here, where the investigators tried to look at the immune sub-population and differentiating them between the micro-environment, and the tumor cells itself.

With this highly granular data also requires approaches, we then be able to understand the different cell clusters and their relationship with the tumor.

We are quite excited that all this should be able to enable researchers to glean better data from some of the new drugs that are trying out in the clinic.

I think this is one fine example of how this technology could be used.

This is a randomized, a very nice single-arm trial whereby the investigators randomized patients to either single agent immune checkpoint inhibitors or combination treatment.

During the course of the treatment itself, patients had to volunteer samples of pre and post-treatment tumor as well as regular blood draws for translational biomarker work.

One of the key findings from this study where the investigators using the technology of single-cell sequencing, they were able to understand that on treatment as the patients receiving these immune checkpoint inhibitors, it actually requires both the immune population of cells proliferating in the tumor microenvironment, as well as cells are circulating in them where this proliferation of cells are equally important because they migrate to the microenvironment.

I think this is a fine example of how all this technology can actually help us understand how these new treatments actually work.

Of course, with all this means to generate data, I think we all appreciate the difficulty and challenge to actually, how do we bring this to the clinic in terms of execution.

I think apart from having a segment in this session to talk about this some of the ideas of bringing the data to clinic, I think we appreciate that there's a big data tsunami with the whole digitization of electric medical health records to now being able to utilize all the data from imaging to molecular profiling, I think, but we all appreciate there's a huge gap in terms of data application.

I think that we acknowledge that decision-making today does not leverage on all the information

I think this is where artificial intelligence really fits in very nicely.

I think this brings me to the other session that we talk about, which is a very nice segue from the multiomics technology where, again, bringing AI across the different domains of healthcare from early detection to analysis of pathological images.

This is really led by the advancement in computer vision to how do we enable AI or bring in AI to help with some of the molecular profiling work, and so as drug development that we have been pursuing over the years.

Finally, we have a speaker that was speak about the cancer and the big data problem.

Again, in healthcare, we see the application of AI across different segments of industry, but in healthcare is really where it's most exciting, where we can see AI being applied across patient data management to image analysis.

In the precision oncology space, both in terms of drug discovery, and interrogation of the molecular data, AI has a big part to play as well.

AI was a big word, really set of talks.

There's a whole breadth of approaches these days on both machine learning, and more recently in deep learning across all these different clinical purposes.

As an indicator of interest, when using PubMed and I just in keyed in the two keywords of AI and cancer, we can see that in a large decade, there's been a huge interest in AI that seems to be growing since in the last five years.

I think this speaks about the whole field, the whole excitement, both in the healthcare as well as the technological industry on the application of AI in healthcare, but just as a cautionary tier.

I think if you look in terms of the research as well as the current market size and the potential revenue forecast, I think we are seeing at least a five-fold increase of projection in terms of revenue.

I think this also speaks to the issue that right now, despite all the efforts in AI, that's been a delay in the implementation process, I think this is multi-factorial both in the clinic as well as in terms of some of the understanding of the technology.

Just as a brief showcase, I'll touch a bit on some of the successes of AI in the different segments of the topic that will be covered at ASCo Breakthrough.

This is one final example whereby in the past when you apply AI technology in the analysis of pathological specimen, you still need the pathologist to actually delineate between a tumor and normal tissue.

This is a very nice work by investigators from the US, where they've now spun on the technology as published in npj Digital Medicine last year, where just by applying computer vision on an entire H and E slide, taking out the manual delineation, they were then able to build a AI score.

Impressively, if you look across the multiple clinical endpoints for predicting metastasis treatment failure and overall survival, using data from clinical trial, they compared that using conventional parameters, which is the SCCI criteria in the grade box.

The AI tool was able to much better predict this outcome across all the endpoints, across a different dataset. This is high-quality datasets from clinical trial. This is certainly very exciting technology here.

Now, in terms of what I shared about the AI, the use of AI in terms of understanding molecular data, this is just one part of it, but this work by investigators whereby we know that conventionally, when you have molecular data, and then you then have to put through well-curated robust computational pipelines to then extract what could be abnormal or mutated in a tumor.

Now this is still very compute and manpower intensive and research led by both groups here, one of them is Dr. Anders Jacobsen, whom we are also inviting to ASCo Breakthrough, who have pioneered techniques that use deep learning to then be able to do very quick screens, and be able to expedite data processing, to then extract the similar information for the clinician in terms of understanding what may be mutated in the tumor.

Finally, in terms of one or the other conventional, and perhaps much more work has been done in terms of use of AI for drug discovery.

I think that we have seen the success of AlphaGo, where AI now is able to predict protein structures very well.

Through the different segments across drug discovery, from target identification, where apart from just predicting the protein structure, the use of NLP, natural language processing, in terms of parsing of literature, also helps to expedite data aggregation from what is out there, to then screening compound libraries, and then ultimately tried to design or shorten the pathway to identifying a hit, and getting a drug into phase 1 trial.

Also in the other space, whereby in terms of drug repurposing, I think there are now AI approaches, which then allows us to not just test single drug to single target, but potentially multiple combination of drugs, whereby we know once we go up to two or three combinations, historical drug high throughput screens, are extremely labor intensive, and we hold the AI could certainly help to shorten the time that is needed for the space. I think, all these are exciting examples of how AI could be used in the clinic. We hope to showcase this at the ASCo Breakthrough meeting.

To conclude, I hope I've given you a good breadth of the whole excitement in terms of new technologies, and how they are penetrating the oncology clinic, not just at the research level, but we've also seen the embracement of the industry partners in terms of bringing this to practice.

I think the idea is that this leads to better treatment and better outcome.

To end, I welcome you all to ASCo Breakthrough meeting, join us, because we're not just covered the science, but also we have the multidisciplinary teams that the audience in one half to actually interact and kicks and brainstorm on all these ideas.

With that, thank you. Now head it back to Peter.

Dr. Peter:

Thank you, Melvin. That was just wonderful.

I just another mention about the breakthroughs.

You know what, one of the things that we recognize, is the vital role of industry in new breakthrough developments.

You start the [unintelligible 01:13:01] presentation.

I think that's the uniqueness of the breakthrough meeting, as we're bringing in basic scientists, translational scientists, computational scientists, from academia, as well as from industry to create a environment for a few days, where we can discuss free think and stimulate each other.

We have some time for a panel discussion. As you know there is a Q&A ability to ask questions. If you wish, please type it in. I see some from Dr. Addy.

Thank you for that. I'm going to start off with a question, Dr. Chua, that I think ties into to some discussions from the other two speakers, which is the question of drug resistance.

We have learned that while we have breakthrough drugs, cancers seemingly have the ability to mutate and develop resistance.

It strikes me what's difference about the genomic areas that we now have some insights into the biological mechanisms of resistance by analyzing pathway and gene expression and protein expression.

We might be able to actually predict development of drug resistance based on the complex NGS sequencing and the multitude of mutations that we frequently see.

That may predict crosstalk or other mechanisms of resistance.

That could lead us to theorize that we could have combinations of drugs to prevent that.

I think that will require a whole different approach to clinical trials to be remade that leap from NGS testing and panels and use of artificial intelligence to understand that complexity, and then get the FDA to approve a trial where we're putting together drugs ad hoc.

Maybe we could talk most of the panel about the drug resistance and how that might lead to different approaches.

Dr. Melvin:

Thanks for the question. Maybe I'll kick off here.

I think in the field is generally things about the whole importance of human resistance.

We see that in the outcomes of most clinical trials, as the tissue then nicely alluded to, in a toll on KRAS D12, whereby although the drugs work, at some point, they essentially stopped working.

I think this speaks to how these days, when we design clinical trials, that we're collecting samples during the intervening periods before the patient actually hits the endpoint.

The idea, again, is that with all these multiomic technologies, that you're collecting the right samples, and these days, we collect blood, as well as tumor if you can, and hopefully, we have enough specimens to allow us to do all the different assays we're going to do.

Now, speaking about liquid biopsy first, I think it's often easy and perhaps more acceptable to the patient to contribute blood samples, to collect blood as a patient undergoing treatment.

In terms, when you think about the liquid biopsy space, there are so many domains that you can now interrogate from the CDC to ctDNA, and then more recently, ctRNA.

Then I think the way, the more complex approaches come in, is you may have one key target or one binary outcome of these assays in terms of editing or treatment.

With the wealth of data that is generated, I think when you think about data integration, it is quite difficult to do it by conventional bioinformatics or even biopsy at its core approaches.

I think that's where AI comes in, where, again, it's still a citizen, an infancy for research, but potentially by the AI could develop a multi-model, to then be able to, number one, validate what we are currently doing conventionally, in terms of treatment go or no-go decision.

I think that's one potential means.

That's why I think, in knowing that this is exciting for your research, that when the investigators now designed the clinical trials is to preempt that potentially there's all this data that can be generated.

We could then try to do it post hoc at this point in time.

I think it could be a few years before we actually get there.

I think what, again, it speaks to the wish list of most of the clinical trials today, where we really want to design a next trial that allows us to adapt treatment real-time, not just on a conventional endpoint of progression of there's no progression.

I think, based on early on information that you derive before the clinical trajectory of disease takes a different turn.

Dr. Peter:

Thank you.

Hello, Lillian, let me rephrase that question for you.

What regulatory processes would need to change that would allow us to look at combinations of targeted therapies, based on sequencing without having to design a clinical trial to show safety and efficacy of combination?

Every time we do FDA, every time we add two drugs, even if they're approved, the FDA feels well, that's risk. Now you got to go back to square one and study the toxicity all over again.

It makes it very difficult to do combination drugs.

When I got molecular tumor boards and they have these are typically patients with advanced cancer.

They tend to have multiple mutations found and there's usually a robust discussion about, "Well, we could target this, we could target that," maybe these two mutations relate to each other, but it's very difficult to go from a very intellectual academic discussion with the genomic scientists, and actually translate that into something we could actually try on a patient.

Dr. Lillian:

Peter, first of all, I don't pretend I'm able to speak for the regulatory agencies since I don't, but I think combination therapy has been the Achilles heels of drug developers and clinicians and scientists all along because it's so difficult to think of using multiple ways to attack the cancer.

That's exactly what we need to do. In most cases, we don't have cancers that depend on one thing. Most cancers depending on multiple ways to survive.

I think there's a lot of interest to say, can we use these modalities that Melvin has talked about to look at the cell states?

What are the states in the cells and the tumor microenvironment to be able to actually select from a variety of drugs to target the cancer and anticipate what it will do in time. I think maybe a new way of thinking is not so trial centric in terms of thinking about a trial, but more patient-specific.

Each patient may have to have his or her own development path that allows us to actually think of how to use these various modalities to predict what works best now, and more importantly, in the later on as it becomes resistant.

That's a complete different paradigm than to think of really a drug or a trial-specific path, which is what we've been following all along all these years in drug development.

We don't really design anything for a single patient, over time but obviously, that's not an easy task.

Dr. Peter:

No, but my sense is that we need to go in that direction because of the complexity of cancer and as we understand the biology better, we need to utilize that.

My other question is for David.

You tantalize us with the concept of going after solid tumors, which brings in the issue of the tumor microenvironment. It's much more complex than liquid tumors, where things are circulating.

The idea of hot and cold tumors where the effector cells do not infiltrate in almost a force field around the cancer blocking the immune system from entry.

That also some way ties into Melvin's discussion about AI visual recognition and single-cell analysis, looking at digital pathology to understand what is happening on a cellular level.

David, as your company or you're thinking about solid tumors, what other challenges do you think you have to overcome that you don't face right now in the liquid tumor arena?

Dr. David:

The issues with solid tumors are many, and we're this one company trying to deal with it. One big area is what are the right targets.

The targets aren't as obvious as they are for hematologic cancers. CD19, we found that you can knock out all the B cells in a person, and people can do quite well.

If you're knocked out, let's say in EGF receptor, or even HER2, they wouldn't.

Finding the right targets on solid tumors, some people are thinking about combinations of targets in order to make that happen. Yes, there's just a lot of work that needs to go on there. It is one thing we're certainly thinking about.

Then, as I mentioned, what edits might you need to do to make the cells work in this difficult environment as you've characterized it?

It comes back to the biology. I think it does come back to a lot of things that Melvin was talking about. How do we understand these tumors on a deeper level in terms of what resistant mechanisms they're putting out to T cells?

What are the specific characteristics that it's not just lung cancer, but the hundred or thousand different types of lung cancers that act in different ways? It is going to have to all come together.

I think what we're looking at first are some very common mechanisms that have been thought to inhibit tumors looking at mechanism, TGF-beta as one of them and some others like that.

I think that's how it's going to advance.

You'll start to identify good targets.

You'll identify some likely targets to enhance the immune response, and we'll move forward to trials.

Hopefully, because it's so rapid that you can edit these things, it does come back to this idea of how I think regulatory agencies are important.

Can we rapidly modify drugs as we move forward to go on to a better version?

It's early days now, I should say we're not, our company isn't quite in the clinic with our technology yet.

We are working with others and other companies who may get there with some of their targets, and we'll get there.

It'll take us a couple of years for the ones that we want to go after the most.

Dr. Peter:

Thank you

 We're actually in the wrap-up session right there so quickly. The time really flew by.

If I could bring up the slides that we wanted to organize. I think there was a question about, is this being recorded and would it be available for others to view or view later?

I'm not actually quite sure of the answer to that, but we will ask the organizers to get back about that question.

With that, I want to thank again our speakers, Dr. Siu, Lebwohl, and Chua and for their time today.

If I have the next slide, I think there's one after that. There's for those of you know, this was purposely set up the week before the J.P. Morgan event in rainy San Francisco.

For those who are in San Francisco this week, there are these three events. These are satellite symposium being organized by each eChinaHealth, and they welcome in-person.

Attendance is not going to be available virtually.

There also will be a social event Monday evening at Pier 9 in San Francisco.

For those who are lucky enough to be in California Monday evening, you're welcome to come by to that event.

Once again, we have our ASCo Breakthrough meeting, Yokohama. I just can't say enough about the meeting. It is ASCO's international meeting.

We purposely at ASCO wanted to go outside the US to go to a different part of the world to disrupt ourselves, as we like to say, and shake things up.

We wanted to find a topic that was different, not an organ-based topic like ASCO GI or ASCO GU, and someone came up with a brilliant idea of novel technology, especially since so much of this is coming from Asia.

This is our second event now that we are moving past the pandemic somewhat to have in-person meetings. This will be in Japan.

Next slide. There will be abstracts that you can submit the presentation. Lillian, I don't remember what the date is of that, but I think we are now accepting abstracts. I think it was just [crosstalk]

Dr. Lillian:

I don't remember either, but we'll get back to you.

Dr. Peter:

I think it just opened, so look out for that.

Once again, we thank Novotech for supporting this evening's activities and then a list of some other events from eChinaHealth. Again, some of the sponsors.

Once again, I thank everybody in the audience. We had over a hundred people participating, and I hope you're all as excited as I was to hear tonight speakers.

Thank you for our speakers and good morning or good evening or wherever you are, and Happy New Year once again. Bye-bye.

Dr. Melvin:

Bye-bye, Lillian

Bye, David.

Dr. Lillian: Bye.

Dr. Peter:

That's right. I think we're done. Thank you. Thank you very much. Happy New Year.

Dr. Melvin:

Bye.