iwMyeloma 2024 | The value of genomics in multiple myeloma & applying this in clinical practice

Hello, my name is Gareth Morgan. I’m professor of hematology at New York University in New York City. I’m here with three of my colleagues today at the International Working Group on Multiple Myeloma. Perhaps you guys could could introduce yourself. Arun.

Okay.

So yeah, Arun Wiita from the University of California, San Francisco, and I’m very happy to be here.

Francesco Maura from University of Miami,

And I’m Sarah Gooding from the University of Oxford.

So today we’ve had a very, very interesting session on the genomics of multiple myeloma and what we try and illustrate is how the talk spread from very basic research on the genome to using it to find new targets, to how you might implement it in the clinic to improve patient outcomes. So from Francesco here, we had a great presentation on classifications. Perhaps start with you, Fran.

Thank you, you know, I think over the last probably almost ten years, we have known more and more about which somatic alteration contributes to the myeloma clinical outcomes and biology. But there are so many and they are often co-occurring one another, which is very hard to really understand which type and how many types of myeloma we have. So the analysis that we performed in collaboration with multiple institutions, I think, you know, has some novelty in how the data were handled.

So we didn’t run only one type of classification, but we run multiple analyses taking into account all the different confounders and co-factor and co-occurrences.

And what was surprising is to just like the simplicity of the outcome is that there are myeloma, that on the top of the canonical translocations or hyperdiploid, which are the, you know, most important events that we know since the 90s, we have two clear patterns: one is like a sort of high genomic complexity with a lot of tumor suppressor genes like TP53, CLD, TRAF3, and another one that is extremely simple – but despite simple, the outcome still can be bad. And you know, there are some mutations like KRAS and RAS that could be therapeutically targeted. And so what we think is that having a more granular classification will allow to better identify patients that can benefit from a certain therapy versus the other, in particular, in a field as rich as multiple myeloma, where most every year.

So tell us about artificial intelligence and how you use that and sort this out.

I think artificial intelligence, the model that we use, was slightly better compared to non-artificial intelligence model. But the real advantage of artificial intelligence will come where we have like 100,000 or 10,000 patients where we it really needs big numbers. So, what we show is that technically it’s possible and practically provide an advantage. But most important, the model based on these types of learning methods are models that allow to predict the risk of each single patient. So right now patients are just assigned as high-risk, low-risk and it’s not really true most of the time how the patient will go; a lot of high-risk go well and a lot of low-risk go actually very bad.

And so the question is can we do better and also can we use the risk to select treatment?

I think these types of deep learning methods are really the answer to that. But because you need the 10,000 100,000 samples, we need to collaborate more between institutions and put together the numbers.

So what’s your take home message?

We need to do more genomics and more collaborations.

Is it going to work?

Absolutely 100%.

Okay, you’ve heard it first from Francesco Maura – genomics is going to work.

But probably he needs also additional, you know, layer of information like proteomics.

So the next step in this process is, you know, envisaged by Arun here, who’s going to tell us a little bit about targeting CAR-T’s to these genomic abnormalities. Right.

So really what we’ve been trying to do in the lab and what we just presented right now is really thinking about now we know so much about genomics due to your work and work of many others really thinking about how do we extract therapeutic targets from this information that we’re getting and it’s incredibly rich data sets.

And the way that we’re trying to think about going that is taking advantage of we have genomic subclassifications in multiple myeloma we know can determine patient risk.

Clearly there’s many degrees and many more layers of subtlety there. But even starting from that standpoint, what we’ve really been focusing on in our lab is t(4;14) myeloma, which many of our audience watching –

So tell us about the 4;14 surfaceome – that was a new word for me today.

So what we really think about is in myeloma, we know that there are thousands of proteins on the surface of tumor cells, essentially,  and we think about existing immunotherapies that are out there – BCMA targeted CAR-T’s or GPRC5D targeted bispecific antibodies. Those are all targeting proteins on the surface of tumor cells.

What we do in my lab, we try and use a technique where we can look at thousands of proteins at the same time, called mass spectrometry. What we can do is we can then take those signatures and try and say, do they correlate with these different genomic subtypes of multiple myeloma. And so in this context, what we really looked at, what we’ve been really focusing on a lot lately is this one protein called CD70. It’s one of these many thousands of proteins that are on the surface of the tumor cells. But what we really found is that it is highly expressed in this t(4;14), high-risk subtype of multiple myeloma, we know these are the patients who tend to relapse the most quickly after BCMA targeted CAR-T. And so these are the patients that we think can we come up with new therapies that target CD70.

So I’m a little bit aware of your work, and so I’m fascinated by this idea of abnormal targets on the cell surface. Can you tell us about that?

Yeah, absolutely. I mean, you know, it’s really interesting because basically we don’t know all the mechanisms, but in many cases, tumor cells do seem to express certain markers that are not there on normal cells. And as opposed to, say, BCMA that’s present on normal plasma cells, they are malignant and still express BCMA in the context of multiple myeloma. Some targets, like CD70, appear to get upregulated, oftentimes we don’t know why yet, but they go to the surface and tumor cells. And so this is something that we’re very interested in my lab is trying to figure out, at least from the outset, saying, what are those targets that are aberrantly expressed on tumor cells that are not on normal tissues that could make good immunotherapeutic targets?

And then we can go back and try and figure out what are the mechanisms that are driving them there, because that could lead to other combination therapy strategies, or just understanding that fundamental biology will help our immunotherapy work better, most likely.

So the last part of this trilogy is Sarah here, and without the work that Sarah’s doing, we have to get these tests to patients in the clinic. And so she gave a great talk describing how they’re trying to do that in the NHS in England. So over to you, Sarah.

Yeah, so what’s interesting in myeloma is we have this fantastic body of research into genetics and risk and how important it is to identify the patients who are going to do less well because their genetics are high risk. And Francesco’s prognostic modeling is what shows us how we identify who those patients are. But all this data is from clinical trials and we have these vast data sets from clinical trials, and that isn’t the real world.

And therefore, the patients we see in clinic, both in specialist centers and in all the regions, their genomes are not being sequenced.

So what we need to do is to take that learning and distill it in a way that our health care funders can afford, and be able to deliver it to everyday patients outside clinical trials.

And that’s what we’re trying to do in the NHS with an easy to use, targeted sequencing panel that takes the main genetic changes in myeloma and delivers them for a low cost that can be delivered across a health care system. And we’ve got a target of 18 month’s time to deliver that in the NHS, and we’re going to do it.

So, you presented some good, I thought examples of how you’re going to validate it by asking clinicians questions. Could you like, tell us about that?

Yeah, so we did a pilot study with 100 patients and we gave clinicians their data as they would receive it at the moment, which is a bit of clinical data and their FISH.

And then we also gave them the same patient but with the genetic data instead.

They didn’t know when they were seeing the same patient and they had to decide how to treat that patient. And what we found was that having that extra genetic information, which enabled them to identify the high-risk patients better, they were able to select the patients that they felt to be high-risk and decide to intensify their treatment, and they were able to do that. I mean, that went up from about 25% of cases to nearly 50% of cases that they were able to decide to intensify. Now, that means that the high-risk patients may get better treatment, but the reverse is probably also true where the patients who don’t have any high-risk disease and Francesco mentioned those, the ones that have the really simple genomes that aren’t becoming more complex, we may be able to de-intensify treatments as well, but that’s another piece of work that needs doing.

So you heard it here from the iwMM 2024 – we really are moving genomics from the laboratory setting to the clinic in two different ways: one is defining new drugs, new treatments against these targets, and one is risk stratifying patients. So soon you’ll be able to have personalized medicine for myeloma patients in the clinic and I’m very hopeful that the work Sarah’s doing is going to get that out there for everybody.

So thank you very much for your attention.