# Aaron Newman: A Breakthrough Blood Test for Cancer

An episode of [Ground Truths](https://invisiblenetwork.ai/podcast/1728526108), 3 Jul 2026, 48 min.

Audio: https://api.substack.com/feed/podcast/199790155/dc1610765ffdaaf6e2514e14bd257944.mp3

The tumor microenvironment (TME), composed of immune cells, fibroblasts, and endothelial cells, has emerged as critical to pathogenesis, prognosis and treatment for cancer (schematic Figure below). There’s no such thing as TMI (too much information) for TME! Yet there’s still no way of assessing it in the clinic by a blood test, and even by invasive biopsy the sampling bias for TME without broader geographic context can lead to important mistakes. That isn’t used for clinical decision making. The TME field is about to change. Recently, Professor Aaron Newman at Stanford and colleagues from many US leading centers published a landmark paper in Nature. It may ultimately be considered one of the most important leaps in cancer diagnostics in decades. Ground Truths is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This was a massive undertaking involving 10 million single-cell RNAseq in 10 types of cancer to define 9 TME types by spatial omics which they called “spatial ecotypes”. It was AI-enabled by machine learning (Spatial Ecotyper from biopsy samples) and extrapolating that to knowledge to deep learning AI (Liquid Ecotyper from blood samples). The whole project took several years. Here’s an infographic summary I made with the help of NotebookLM along with the Figures that follow. Some of the key points from our conversation: —TME is the soil in which the cancer cells live. It strongly influences whether the cancer cells will thrive or not, spread, respond to therapy, and determine the fate pf pre-cancerous lesion (as seen in recent important paper in pancreatic cancer evolution and colon cancer development). —Ecotypes, a term coined in 1922, to get at the local environment of tissue and cancer, was repurposed in this new project. Good metaphor: neighborhoods. And emphasis on location , location, location like real estate. —In the clinic today, the only window to TME is via a biopsy. But decisions for treatment are not based on TME, they are based on biomarkers like tumor mutation burden (TMB), PD-L1 levels , or volume of tumor DNA by liquid biopsy. —TME is dynamic and changes over time so a single snapshot won’t cut it. —Spatial 3D maps of TME from 10 million cells, 10 cancer types led to 9 spatial ecotypes (with the Spatial Ecotyper machine learning). Validation across 3D grids (such as Visium) and multiple cohorts. The cell location base within TME is striking. —These 9 neighborhoods are conserved and very different. SE4-immune system is asleep, hypoxic; SE 7/8 are “war zones” rich with immune cells, hot tumor; SE 5 exhibits intense immunosuppression, hostile vs immune system, portends poor survival, SE9 deep in core, driving angiogenesis (new blood vessels) —Now rich information from these 9 SEs needs to be extrapolated to a holistic blood test of TME. It turns out an extension of tapping into methylation as used by some liquid biopsy cell-free DNA tests works very well, providing cell identity from the DNA fragments leaked into the blood. To get to TME spatial biology in a tube of blood. Akin to wastewater surveillance form municipals for detection of infectious disease pathogen sequence variants and circulating levels. A triumph of AI analytics made this possible! —The Liquid Ecotyper was validated in 78 patients with malignant melanoma pre-treatment from Yale University. SE 7/8 benefited from immune checkpoint therapy; SE4 resisted treatment, shorter survival. Matched with tissue biopsy work. —Work beyond publication ongoing in many types of lung cancer, bladder cancer, mesothelioma, and response to other therapies like CAR T and bispecific antibodies. This is critical since we have slew of different cancer immunotherapies (see my pyramid Ground Truths review) and we’re not sure who, when, how to use them to drive optimal cancer outcomes in patients. May also help predict other types of therapy (not just immunotherapy) for cancer.…

## Transcript

[0:00] SPEAKER_02: I think we got a quorum.
[0:02] SPEAKER_02: We'll have more, of course, as we get going here.
[0:08] SPEAKER_02: So I am thrilled to welcome Professor Aaron Newman from Stanford,
[0:14] who published a paper that will probably be looked at as a
[0:18] classic for years to
[0:19] SPEAKER_02: come,
[0:20] SPEAKER_02: where he broke down this tumor microenvironment,
[0:25] SPEAKER_02: also known as TME,
[0:27] SPEAKER_02: to be able to have a blood test,
[0:30] a diagnostic test that can follow the tumor microenvironment in
[0:37] a person's cancer.
[0:38] SPEAKER_02: This has extraordinary implications.
[0:40] SPEAKER_02: So, Aaron, welcome to Ground Truths.
[0:45] SPEAKER_00: Thanks, Eric, for having me.
[0:46] SPEAKER_02: Yeah.
[0:47] SPEAKER_02: Well, I want to congratulate you.
[0:49] SPEAKER_02: I think this reflects many,
[0:51] SPEAKER_02: many years of work finally harvested in this extraordinary paper,
[0:56] SPEAKER_02: which when this gets fully posted,
[0:59] SPEAKER_02: we'll link to,
[0:59] SPEAKER_02: of course.
[1:00] SPEAKER_02: But
[1:01] SPEAKER_02: I thought what we would do is start with the importance,
[1:05] the significance of this tumor microenvironment because with
[1:10] every week that goes
[1:11] SPEAKER_02: by,
[1:11] SPEAKER_02: especially right now or right after the ASCO meetings,
[1:15] SPEAKER_02: this TME gets even more prominent in its importance.
[1:19] SPEAKER_02: So what is TME?
[1:22] Yeah,
[1:23] so the tumor microenvironment is in many ways the soil in which
[1:27] cancer cells live.
[1:29] SPEAKER_00: And it's a context.
[1:31] It's an environment that includes genetically normal cells of
[1:35] the immune system,
[1:36] as well as...
[1:38] SPEAKER_00: Stromal cells,
[1:39] which are cells that are structural cells like endothelial
[1:43] cells that produce blood
[1:44] SPEAKER_00: vessels and fibroblasts that produce scaffolding out of collagen,
[1:48] SPEAKER_00: etc.
[1:49] But in a nutshell,
[1:49] it's the genetically normal cells that cancer cells find
[1:52] themselves in.
[1:53] SPEAKER_00: And this soil is incredibly important because it,
[1:57] SPEAKER_00: in many ways,
[1:58] SPEAKER_00: will strongly influence whether a cancer cell will thrive or not.
[2:03] SPEAKER_00: You know,
[2:03] SPEAKER_00: you can think of this as,
[2:05] SPEAKER_00: you know,
[2:05] cancer cells are weeds in a garden and the soil will determine
[2:08] whether this
[2:08] whether the weeds are going to grow or they're going to wither,
[2:11] whether the cancer cells might
[2:14] SPEAKER_00: You know,
[2:14] spread or metastasize to neighboring lymph nodes or to distal
[2:18] sites in the body in
[2:19] SPEAKER_00: the same way that weeds might spread.
[2:21] SPEAKER_00: But also how they respond to therapy to the extent of that response.
[2:27] And increasing evidence that suggests that this tumor
[2:30] microenvironment soil can
[2:31] SPEAKER_00: also strongly determine the propensity for early cancer,
[2:36] SPEAKER_00: like precancerous lesions,
[2:38] SPEAKER_00: to become invasive malignancy.
[2:41] SPEAKER_00: So it is incredibly important to
[2:44] SPEAKER_00: And,
[2:44] SPEAKER_00: you know,
[2:45] SPEAKER_00: not only that,
[2:46] but this relationship between sort of the local environment and
[2:50] the characteristics
[2:51] SPEAKER_00: of,
[2:51] SPEAKER_00: let's say,
[2:52] SPEAKER_00: a weed,
[2:53] SPEAKER_00: you know,
[2:53] SPEAKER_00: or other kinds of plant species or even,
[2:55] SPEAKER_00: you know,
[2:55] SPEAKER_00: any species,
[2:57] SPEAKER_00: it actually has a pretty long history.
[2:58] SPEAKER_00: It can be traced back to,
[3:00] SPEAKER_00: you know,
[3:00] SPEAKER_00: over 100 years ago,
[3:01] SPEAKER_00: this Swedish evolutionary botanist in 1922 coined the term ecotype.
[3:08] which actually sounds like a like a modern term but it actually
[3:11] is a way to
[3:12] understand how the characteristics or the physical traits of a
[3:16] species depend on
[3:18] its local environment and when we learned about that we're like
[3:21] you know aha you
[3:22] SPEAKER_00: know that's a really
[3:24] Great concept and a strong framework for thinking about how
[3:28] cells in any tissue,
[3:29] SPEAKER_00: including in cancer,
[3:30] SPEAKER_00: and in particular the tumor microenvironment,
[3:34] SPEAKER_00: can change as a function of their context,
[3:38] SPEAKER_00: where the cells are in space,
[3:40] SPEAKER_00: their physical location,
[3:41] SPEAKER_00: either inside of the tumor or on the edge of the tumor or beyond it.
[3:46] And so we essentially repurposed that concept and we use it to
[3:50] understand and
[3:51] SPEAKER_00: describe the tumor microenvironment in cancer and its different context
[3:57] SPEAKER_00: dependencies,
[3:58] SPEAKER_00: its relationships to clinical outcomes.
[4:01] It's been an incredibly useful framework for us and we've been
[4:05] using this to try to
[4:06] SPEAKER_00: understand,
[4:07] SPEAKER_00: you know,
[4:07] SPEAKER_00: are there
[4:08] Ecotypes of the tumor microenvironment that are shared across
[4:11] different tumor
[4:12] samples or different cancer types do they differ depending on
[4:16] for example the cells
[4:18] SPEAKER_00: that are inside of the tumor you can think about them as you know maybe
[4:22] The conversations that are happening inside City Hall and like
[4:25] an urban environment
[4:26] will be potentially very different than the conversations in
[4:29] PTA meetings in the
[4:31] SPEAKER_00: suburbs,
[4:31] SPEAKER_00: which we can think of as maybe the edge of the tumor.
[4:34] SPEAKER_00: And those will be different from,
[4:35] SPEAKER_00: you know,
[4:36] ecotypes or conversations that are further away in what we call
[4:39] the outer stroma.
[4:41] SPEAKER_00: Maybe in our city analogy,
[4:43] the rural communities talking about agricultural methods and
[4:46] things like that.
[4:46] SPEAKER_00: So,
[4:47] SPEAKER_00: you know,
[4:48] SPEAKER_00: the location matters,
[4:49] SPEAKER_00: the neighborhood matters,
[4:51] SPEAKER_00: the cells that they talk to,
[4:53] SPEAKER_00: their social networks,
[4:54] all of these things are sort of encompassed in this concept of
[5:00] a tumor ecotype.
[5:01] SPEAKER_02: Yeah,
[5:02] SPEAKER_02: I love your metaphor of the soil for cancer,
[5:06] SPEAKER_02: which could try to repress,
[5:08] SPEAKER_02: suppress it,
[5:08] SPEAKER_02: or could actually provide more nutrient blood supply and let it flourish.
[5:18] SPEAKER_02: Now, the basic constituents, immune cells, fibroblasts, matrix cells,
[5:28] SPEAKER_02: I think,
[5:29] SPEAKER_02: you know,
[5:29] SPEAKER_02: we've kind of,
[5:30] SPEAKER_02: the gamut is,
[5:33] SPEAKER_02: you know,
[5:33] a bunch of things that are in this stuff said in a recent paper
[5:40] showed that we have
[5:43] SPEAKER_02: hundreds if not thousands of these so-called panans,
[5:45] SPEAKER_02: each of us as we get older,
[5:47] SPEAKER_02: in our pancreas,
[5:48] SPEAKER_02: these little precancerous lesions.
[5:51] SPEAKER_02: And just published was the fact that it's
[5:53] the microenvironment that leads some of these to actually
[5:56] develop pancreatic cancer.
[5:58] SPEAKER_02: So we keep learning,
[6:00] SPEAKER_02: and that was just last week,
[6:01] SPEAKER_02: we keep learning more and more the significance of this TME.
[6:07] SPEAKER_02: Now in clinical practice today, before we get to your breakthrough work,
[6:12] SPEAKER_02: Is the TME assessed in patients with cancer?
[6:16] Because that requires a biopsy,
[6:18] which would be a sampling bias,
[6:20] of course.
[6:22] SPEAKER_02: And you couldn't do that every X number of months.
[6:26] But is this part of the way a person would have their cancer
[6:30] ever assessed in a clinic?
[6:33] Yeah,
[6:33] so in the clinic,
[6:35] you know,
[6:36] patients may obtain at most in general a diagnostic biopsy.
[6:40] SPEAKER_00: You know, biopsies are risky and a lot of patients will receive them.
[6:45] But if there is a biopsy,
[6:47] there are ways to access the tumor microenvironment.
[6:50] SPEAKER_00: But generally speaking,
[6:51] SPEAKER_00: and especially when it comes to therapy decisions in the clinic,
[6:54] SPEAKER_00: most therapy decisions are cancer cell-centric.
[6:58] SPEAKER_00: They're not generally looking at the tumor microenvironment.
[7:01] SPEAKER_00: The exception of that would be to look at,
[7:04] SPEAKER_00: for example,
[7:05] SPEAKER_00: a protein called PD-L1,
[7:07] SPEAKER_00: which is associated with response to immune checkpoint inhibitors,
[7:11] SPEAKER_00: but also has its problems.
[7:13] SPEAKER_00: But in general,
[7:14] SPEAKER_00: no,
[7:14] the tumor microenvironment is not used to make clinical
[7:17] decisions outside of that.
[7:20] more importantly you know the tumor microenvironment like the
[7:23] cancer itself is not
[7:24] a static entity it changes over time and if we only have that
[7:29] single biopsy to use
[7:32] to make a decision about let's say the therapy that a patient
[7:34] might respond to
[7:36] SPEAKER_00: Eric Topol erictopol
[8:00] SPEAKER_00: In response to therapy,
[8:01] SPEAKER_00: you know,
[8:02] we know that the therapies that patients receive will likely
[8:05] remodel the tumor
[8:06] SPEAKER_00: microenvironments for new ways,
[8:08] SPEAKER_00: but we have no clinical assay to access that information over time.
[8:12] SPEAKER_00: It just simply doesn't exist.
[8:14] SPEAKER_02: Yeah.
[8:16] SPEAKER_02: Well,
[8:16] so this is really important because what you've reviewed
[8:19] already is that it is rich
[8:22] SPEAKER_02: information in this
[8:24] SPEAKER_02: microenvironment.
[8:25] SPEAKER_02: And by the way,
[8:26] I should have mentioned endothelial cells are the seat for
[8:29] blood vessels as the
[8:30] SPEAKER_02: other constituent in the TME.
[8:33] SPEAKER_02: But it's dynamic so that if you do take a biopsy,
[8:37] SPEAKER_02: not only are you getting that specific one little piece,
[8:40] SPEAKER_02: but you also are not able to project over time how it's going to change.
[8:45] SPEAKER_02: So it isn't used on it.
[8:47] SPEAKER_02: And of course, the dream would be to have a blood test.
[8:53] SPEAKER_02: A blood test that would tell us at any moment in time what was going on
[8:57] SPEAKER_02: holistically in a person's cancer,
[8:59] SPEAKER_02: tumor,
[9:00] SPEAKER_02: and microenvironment.
[9:01] SPEAKER_02: So that mission,
[9:03] SPEAKER_02: which you,
[9:04] SPEAKER_02: you know,
[9:04] SPEAKER_02: that kind of reminds me of Mr.
[9:06] SPEAKER_02: Phelps in Mission Impossible,
[9:08] SPEAKER_02: you know.
[9:08] You're going to assemble all these medical centers around the
[9:12] country and you're
[9:13] SPEAKER_02: going to go on this multi-year campaign to go after this.
[9:16] SPEAKER_02: I mean, it's just incredible.
[9:19] SPEAKER_02: So the first phase of this is a solid phase.
[9:23] SPEAKER_02: You're going to get all these tumors,
[9:25] SPEAKER_02: 10 million cells from 10 types of cancer,
[9:29] SPEAKER_02: and you're going to get spatial 3D maps.
[9:34] SPEAKER_02: If I could sum it up,
[9:37] spatial transcriptomics of the gene expression for 10 million
[9:41] cells in these 10
[9:42] SPEAKER_02: cancer types to basically get the map of the tumor microenvironment,
[9:47] SPEAKER_02: knowing that it's location,
[9:49] SPEAKER_02: location,
[9:49] SPEAKER_02: location,
[9:50] SPEAKER_02: like real estate,
[9:52] SPEAKER_02: and
[9:53] SPEAKER_02: That there's these neighborhoods,
[9:54] SPEAKER_02: as you said,
[9:55] SPEAKER_02: could be in the core of the cancer,
[9:57] SPEAKER_02: could be in the margin of the cancer,
[9:59] SPEAKER_02: could be at the healthy tissue,
[10:00] SPEAKER_02: it could be,
[10:02] SPEAKER_02: you know,
[10:02] SPEAKER_02: various different places to define ecotypes,
[10:05] SPEAKER_02: or maybe could you use the word neighborhoods?
[10:09] SPEAKER_02: Okay,
[10:10] so tell us about the first machine learning of all these
[10:14] massive number of cells to
[10:16] SPEAKER_02: get these spatial nine ecotypes.
[10:21] SPEAKER_00: So,
[10:22] SPEAKER_00: you know,
[10:22] SPEAKER_00: the scale of the problem really necessitated a computational solution,
[10:27] SPEAKER_00: a machine learning based solution.
[10:28] It's just overwhelming the amount of information that these
[10:32] modern technologies can generate.
[10:34] SPEAKER_00: So spatial transcriptomics, as you mentioned,
[10:36] and this is a technology that allows us to determine which
[10:40] genes are expressed in
[10:41] SPEAKER_00: individual cells and also their spatial coordinates.
[10:44] SPEAKER_00: So we know,
[10:45] SPEAKER_00: for example,
[10:46] SPEAKER_00: in the tumor microenvironment where specific types of T cells are,
[10:49] SPEAKER_00: the endothelial cells,
[10:51] SPEAKER_00: the fibroblasts,
[10:51] SPEAKER_00: etc.,
[10:52] SPEAKER_00: and where the cancer cells are in relation to them.
[10:56] SPEAKER_00: And then we have, you know, samples from different
[10:58] SPEAKER_00: patient tumors from different cancer types.
[11:01] SPEAKER_00: And one of the fundamental questions has been,
[11:03] are there commonalities in these neighborhoods or tumor
[11:07] microenvironments across
[11:08] SPEAKER_00: patients,
[11:09] SPEAKER_00: across cancer types?
[11:10] SPEAKER_00: Is there sort of a fundamental parts list that is used to reconstruct,
[11:15] SPEAKER_00: to create the tumor microenvironment across patients?
[11:17] SPEAKER_00: So we needed to use machine learning to really answer that question.
[11:22] SPEAKER_00: And so we developed a technique called spatial ecotyper,
[11:25] SPEAKER_00: And a spatial ecotype,
[11:26] SPEAKER_00: essentially what it does is it takes all of that information,
[11:30] SPEAKER_00: those individual gene expression measurements of cells,
[11:33] SPEAKER_00: the spatial locations within each of these tumor samples,
[11:37] and it integrates it all into what we call an embedding or a
[11:41] space that we can then
[11:42] SPEAKER_00: ask,
[11:43] SPEAKER_00: you know,
[11:43] SPEAKER_00: are there commonalities?
[11:44] Are there specific neighborhoods that tend to recur across
[11:47] tumor samples and across
[11:49] SPEAKER_00: cancer types?
[11:50] SPEAKER_00: And when we did this and we sort of
[11:53] SPEAKER_00: Eric Topol erictopol
[12:08] SPEAKER_00: These are actually cancer types with distinct cell of origin,
[12:12] SPEAKER_00: but they share remarkably these different tumor microenvironments.
[12:27] We then went on to validate the presence of these spatial
[12:32] ecotypes across different
[12:34] SPEAKER_00: spatial transcriptomics modalities,
[12:36] SPEAKER_00: across different cohorts of patients,
[12:39] SPEAKER_00: and ultimately looking at 10 different cancer types,
[12:43] SPEAKER_00: 9 carcinomas and melanomas.
[12:45] So this was really striking to us,
[12:48] how conserved these microenvironments are.
[12:51] SPEAKER_00: We were able to even...
[12:53] SPEAKER_00: Identify them,
[12:54] SPEAKER_00: you know,
[12:54] SPEAKER_00: once we knew about them in non-spatial data,
[12:57] so single-cell gene expression data that had been generated
[13:00] over the years by many
[13:01] SPEAKER_00: different groups,
[13:03] as another form of validation that these spatial ecotypes and
[13:07] their constituent
[13:08] tumor microenvironment cell types could actually be
[13:13] re-identified in these
[13:15] SPEAKER_00: independent data sets.
[13:17] Now,
[13:17] a fundamental feature of a spatial ecotype is that it is
[13:20] comprised of
[13:21] SPEAKER_00: Not only different cell types within a social network or community,
[13:26] SPEAKER_00: but they have unique states.
[13:28] SPEAKER_00: So for example,
[13:29] SPEAKER_00: you know,
[13:30] SPEAKER_00: different gene expression programs of a CD8 T cell,
[13:34] SPEAKER_00: we will see that depending on the spatial ecotype that it's present in.
[13:37] And what that means is that,
[13:39] you know,
[13:39] the biological characteristics differ.
[13:42] You know,
[13:42] the conversations,
[13:43] the social networks differ depending on the spatial ecotype.
[13:47] SPEAKER_02: I thought that was fascinating.
[13:48] SPEAKER_02: When you have a CD8 cytotoxic cell in the tumor core,
[13:55] SPEAKER_02: it would have one behavior,
[13:57] SPEAKER_02: whereas if it was in the margin,
[13:58] SPEAKER_02: it had a totally different behavior,
[13:59] SPEAKER_02: just to show how incredibly dynamic and location-based this really is.
[14:04] SPEAKER_02: I mean, it's so striking.
[14:07] SPEAKER_00: Absolutely.
[14:08] SPEAKER_00: Wow.
[14:08] SPEAKER_02: Yeah.
[14:10] And we identified that there were distinct gene expression
[14:13] programs depending upon
[14:14] SPEAKER_00: the cell type,
[14:15] SPEAKER_00: and we expected to see that.
[14:16] SPEAKER_00: What we didn't expect to find was that there were also...
[14:20] programs or biological processes that were largely independent
[14:23] of the cell type,
[14:24] SPEAKER_00: but that were shared among all the cells within the spatial ecotype.
[14:28] SPEAKER_00: So you can think about it as,
[14:29] SPEAKER_00: you know,
[14:30] SPEAKER_00: people sitting around a table having a conversation,
[14:32] and that conversation is independent of the characteristics of
[14:35] each person,
[14:35] SPEAKER_00: you know,
[14:36] SPEAKER_00: their height,
[14:36] SPEAKER_00: their hair color,
[14:37] SPEAKER_00: etc.
[14:38] SPEAKER_00: So, and then the location matters as well.
[14:42] SPEAKER_00: And so,
[14:43] SPEAKER_00: Eric Topol erictopol
[14:58] SPEAKER_00: Dr. Justin Marchegiani
[15:14] the tumor itself is sort of a you know like an insult that is
[15:17] sensed by these by
[15:18] these cells in the surroundings and they're sort of responding
[15:21] to that and
[15:21] undergoing an acute stress response there are specific ecotypes
[15:26] that are associated
[15:26] with for example wound healing which is another response to the
[15:30] presence of cancer
[15:31] and also interferon signaling which is well known to be
[15:34] associated with you know
[15:35] SPEAKER_00: and it
[15:37] SPEAKER_00: cells that can attack the cancer cells, so immune cells, etc.
[15:40] And there are different interferon signaling programs that are
[15:43] spatially distinct.
[15:45] SPEAKER_00: So there's regionalization,
[15:47] and that ends up being very important in the associations
[15:50] between spatial ecotypes
[15:52] SPEAKER_00: and clinical outcomes.
[15:54] Yeah,
[15:54] so by the way,
[15:55] any of you all,
[15:57] nearly a few hundred are joined already.
[15:59] If you have any questions for Aaron,
[16:02] don't hesitate to put that in the message.
[16:06] I do want to just point out of these nine spatial ecotypes or
[16:11] neighborhoods or
[16:11] SPEAKER_02: whatever you want to call them,
[16:12] SPEAKER_02: if people are afraid of that term,
[16:14] SPEAKER_02: spatial ecotypes.
[16:16] SPEAKER_02: Basically, it's fascinating because, for example, we'll get back to SE4.
[16:22] SPEAKER_02: The immune system is asleep.
[16:24] SPEAKER_02: It's a low oxygen story.
[16:27] SPEAKER_02: But on the other hand,
[16:28] SPEAKER_02: as mentioned,
[16:29] SPEAKER_02: Aaron mentioned,
[16:30] SPEAKER_02: SE7 and 8,
[16:31] SPEAKER_02: those two ecotypes,
[16:33] SPEAKER_02: they're like war zones.
[16:35] SPEAKER_02: Lots of interferon.
[16:36] SPEAKER_02: The immune system is revved up.
[16:39] SPEAKER_02: That's a good thing.
[16:40] SPEAKER_02: That's what you want.
[16:41] SPEAKER_02: Talk about, we want hot tumors.
[16:43] SPEAKER_02: We don't want cold tumors.
[16:44] SPEAKER_02: And so SE4, that acetype would be a cold tumor.
[16:49] SPEAKER_02: And SE5, also an intense immunosuppression.
[16:54] SPEAKER_02: It's kind of hostile against the immune system.
[16:56] You don't want that if you're going to get a treatment that's
[16:59] going to work in your cancer.
[17:00] So these are providing,
[17:02] in the solid analysis with all the machine learning,
[17:06] SPEAKER_02: and all this brilliant work.
[17:08] They're providing new insights we've never had before as what
[17:11] you said conserve
[17:13] SPEAKER_02: across all these different cancer types,
[17:15] SPEAKER_02: these TME,
[17:19] SPEAKER_02: the story,
[17:20] SPEAKER_02: the story is unfolding.
[17:22] So there's so many aspects of this that are novel that we
[17:26] didn't know that the TME
[17:28] SPEAKER_02: would be conserved across the
[17:31] many different cancers and that you'd be able to spatially map
[17:35] this using AI.
[17:37] SPEAKER_02: You're using all the tools of 2026 here.
[17:39] That's what we're going to get to is the deep learning with the
[17:44] liquid ecotyper.
[17:47] SPEAKER_02: So it wasn't enough that you did the solid mapping,
[17:51] SPEAKER_02: but now you're trying to say,
[17:54] SPEAKER_02: okay,
[17:54] SPEAKER_02: we want to get that information and get it so that in a tube of blood,
[18:00] SPEAKER_02: we can find which one of these ecotypes is present in what extent.
[18:06] So how do you go from this rich spatial information to being
[18:11] able to go to this
[18:12] SPEAKER_02: deep learning AI liquid ecotyper?
[18:17] SPEAKER_00: So that's a really great question,
[18:18] SPEAKER_00: and this was really motivated by our findings that,
[18:21] SPEAKER_00: you know,
[18:22] SPEAKER_00: different spatial ecotypes,
[18:23] SPEAKER_00: as you mentioned,
[18:24] SPEAKER_00: SE4,
[18:24] SPEAKER_00: SE7,
[18:25] SPEAKER_00: SE8,
[18:26] SPEAKER_00: etc.,
[18:26] SPEAKER_00: are strongly associated with clinical outcomes.
[18:29] SPEAKER_00: Sort of a hot tumor like SE7,
[18:31] SPEAKER_00: SE8,
[18:32] we found to be associated with response to a specific type of
[18:35] immunotherapy called
[18:37] SPEAKER_00: immune checkpoint inhibitors,
[18:38] SPEAKER_00: which you recently wrote about in your excellent essay on the subject.
[18:42] SPEAKER_00: SE4 associated with resistance to immunotherapy.
[18:45] So we observed that there are these distinct clinical
[18:47] associations for different ecotypes.
[18:50] And the question is,
[18:50] how do you then take such a finding and make that clinically
[18:53] accessible?
[18:53] SPEAKER_00: How do you make that practical?
[18:55] SPEAKER_00: So it turns out that in the blood,
[18:57] there are lots of different analytes that one can access that
[19:01] are revealing about
[19:02] SPEAKER_00: different physiological and pathologic states.
[19:06] SPEAKER_00: And one such analyte
[19:09] SPEAKER_00: This is DNA that is naturally fragmented,
[19:12] SPEAKER_00: that is released from cells throughout the body,
[19:15] SPEAKER_00: mostly healthy cells,
[19:16] SPEAKER_00: but also cancer cells or TME cells.
[19:20] This is essentially a large number of DNA fragments that one
[19:23] can access just by
[19:24] simply taking a blood sample and removing and sequencing the
[19:28] cell-free DNA.
[19:29] Now,
[19:29] historically,
[19:30] cell-free DNA has been used in clinical oncology to measure
[19:34] cancer cells.
[19:35] SPEAKER_00: It's a lot easier because we can look at,
[19:37] SPEAKER_00: for example,
[19:38] SPEAKER_00: mutations,
[19:39] somatic mutations that are not present in the normal cells that
[19:43] are a hallmark of
[19:43] SPEAKER_00: the cancer cells.
[19:45] SPEAKER_00: The tumor microenvironment is different,
[19:46] SPEAKER_00: though,
[19:47] SPEAKER_00: because,
[19:47] SPEAKER_00: as I mentioned,
[19:47] SPEAKER_00: it's genetically normal.
[19:49] SPEAKER_00: We do not have those mutations that we can use as a handle.
[19:53] But there are other features that we can use in an area called
[19:56] epigenomics.
[19:57] SPEAKER_00: There are these chemical modifications called methyl groups.
[20:01] And methyl groups are very specific to cell identity,
[20:05] you know,
[20:05] what type of cell it is.
[20:07] Different cell types have different constellations of these
[20:10] methyl groups.
[20:11] SPEAKER_00: And this is also true for different states of a cell type.
[20:14] And so,
[20:15] you know,
[20:16] the realization here was that,
[20:17] well,
[20:17] if we could somehow lift over
[20:20] SPEAKER_00: What we learned from spatial transcriptomics into this methylation space,
[20:25] SPEAKER_00: then perhaps using AI as the way to do it,
[20:29] we could then measure the levels of these tumor
[20:31] microenvironment neighborhoods
[20:33] SPEAKER_00: noninvasively from the blood.
[20:34] SPEAKER_00: And that's exactly what we did.
[20:37] SPEAKER_00: And this was completely dependent on having an AI-based approach.
[20:41] SPEAKER_00: This is very complicated data.
[20:43] Lots of potential methylation sites around,
[20:46] you know,
[20:46] the genome to learn these signals.
[20:48] SPEAKER_00: But we hypothesized that this, in fact, could work.
[20:51] SPEAKER_00: And so we went ahead and tested that.
[20:54] SPEAKER_02: I'm sorry.
[20:55] SPEAKER_02: But just to differentiate,
[20:57] SPEAKER_02: companies like Grail,
[20:59] they do methylation to basically determine a cancer is present
[21:04] and using AI to
[21:06] That is what you're getting at is that the methylation marks
[21:09] are really rich with
[21:10] SPEAKER_02: the help of AI that there's a lot of
[21:22] Who would have ever thought that we could get the spatial
[21:25] ecotypes from the
[21:26] SPEAKER_02: methylation of the DNA?
[21:28] SPEAKER_02: Because,
[21:29] SPEAKER_02: hey,
[21:30] it hasn't been that long that we've been getting the mutations
[21:33] through the
[21:34] SPEAKER_02: epigenetic methylation mark.
[21:37] SPEAKER_02: So were you surprised you could get that from this?
[21:40] SPEAKER_02: I mean, methylation is already out there being used.
[21:43] SPEAKER_02: It's been used in more than half a million people through DNA.
[21:47] SPEAKER_02: Grail and other types of these liquid biopsy cancer efforts.
[21:53] SPEAKER_00: Yeah,
[21:53] SPEAKER_00: so,
[21:54] SPEAKER_00: you know,
[21:54] SPEAKER_00: we,
[21:55] SPEAKER_00: so starting with the hypothesis that,
[21:57] SPEAKER_00: you know,
[21:57] SPEAKER_00: cell-free DNA is likely being released from the tumor microenvironment,
[22:00] SPEAKER_00: this is something that we hypothesized many years ago,
[22:03] SPEAKER_00: but then we also,
[22:04] SPEAKER_00: you know,
[22:05] as single-cell RNA sequencing was coming online and people
[22:10] started generating these
[22:11] large cell atlases of cancer samples that included the tumor
[22:16] microenvironment,
[22:16] SPEAKER_00: but it also included sort of the healthy immune cells in the blood,
[22:21] SPEAKER_00: It became clear that there were differences,
[22:23] SPEAKER_00: that there were differences in the microenvironment.
[22:26] A CDAT cell that is inside of a tumor is different in terms of
[22:29] its gene expression
[22:30] SPEAKER_00: program than it is further away from the tumor or in the blood.
[22:34] SPEAKER_00: And then if that's the case,
[22:36] then the underlying machinery that leads to those gene
[22:40] expression differences must
[22:41] SPEAKER_00: also be different.
[22:42] SPEAKER_00: That's sort of the logical conclusion.
[22:45] SPEAKER_00: And so the methylation controls gene expression.
[22:48] It's one of the many different features in cells that's used to
[22:52] control how genes are expressed.
[22:55] So we would then hypothesize that if we could use AI to really
[23:00] sift through all of
[23:01] SPEAKER_00: this data that perhaps we could identify
[23:04] Well,
[23:04] this was a coup to be able to take the solid information that
[23:06] you generated and
[23:32] SPEAKER_02: and then basically get the signal or the
[23:36] SPEAKER_02: The Postal Address or whatever you want to call it,
[23:38] SPEAKER_02: the return address for the ecotypes in the blood.
[23:41] SPEAKER_02: I mean,
[23:41] this was yet another dimension of how this came up with
[23:46] something I don't know
[23:47] SPEAKER_02: anyone could have envisioned that you'd be able to do this.
[23:51] SPEAKER_02: So now,
[23:52] with this liquid ecotyper having found this signature for the
[23:57] spatial ecotypes,
[23:59] SPEAKER_02: the solid ones now in liquid form,
[24:03] You then took like 78 or so people with melanoma and you were
[24:08] able to get the Zika
[24:10] SPEAKER_02: types and show the prognostic importance,
[24:13] SPEAKER_02: the relation to therapy,
[24:15] SPEAKER_02: immunotherapy.
[24:16] SPEAKER_02: Pretty remarkable validation.
[24:19] SPEAKER_02: Maybe you want to comment on the melanoma experience.
[24:23] SPEAKER_00: Yeah, definitely.
[24:25] We first started off looking at the spatial ecotypes in sort of
[24:28] the solid tissue
[24:30] SPEAKER_00: context,
[24:30] SPEAKER_00: and that included,
[24:31] SPEAKER_00: you know,
[24:31] SPEAKER_00: melanomas,
[24:32] SPEAKER_00: non-swald lung cancer,
[24:33] SPEAKER_00: bladder cancer.
[24:35] And so this was just looking at a lot of different RNA
[24:38] sequencing data from solid tissues.
[24:40] And we identified a very strong association between specific
[24:44] spatial ecotypes,
[24:45] SPEAKER_00: like I mentioned earlier,
[24:46] SPEAKER_00: like 7 and 8.
[24:47] SPEAKER_00: were associated with benefit to immunotherapy,
[24:49] SPEAKER_00: including in melanoma,
[24:51] SPEAKER_00: an SE4,
[24:51] SPEAKER_00: which is a wound healing based margin associated,
[24:55] SPEAKER_00: ecotype associated with resistance.
[24:57] SPEAKER_00: So we wanted to test whether the same signal could be extracted from
[25:02] SPEAKER_00: Eric Topol erictopol Ground Truths
[25:26] And when we looked at the association of each spatial ecotype
[25:29] with their response
[25:31] SPEAKER_00: to immunotherapy,
[25:32] SPEAKER_00: it was very striking.
[25:34] SPEAKER_00: A nearly perfect correlation between,
[25:37] SPEAKER_00: you know,
[25:38] the individual spatial ecotype response associations in the
[25:42] blood test and those
[25:43] that we had also observed in independent melanoma patients from
[25:46] their surgical
[25:48] SPEAKER_00: tumor biopsies.
[25:49] So that indicated that we could in fact use a blood-based assay
[25:54] to interrogate,
[25:55] SPEAKER_00: you know,
[25:55] the tumor microenvironment and the clinical significance of
[25:59] these different spatial
[26:00] SPEAKER_00: microenvironments.
[26:01] SPEAKER_02: Yeah, that was so striking.
[26:04] SPEAKER_02: This accomplished many things,
[26:05] SPEAKER_02: but for one,
[26:06] you had the tissue to show that your liquid methylation was
[26:11] matching up so nicely
[26:13] SPEAKER_02: with tissue,
[26:14] SPEAKER_02: but you also were able to show that you could get,
[26:17] SPEAKER_02: for example,
[26:18] SPEAKER_02: the ecotype 4 or ecotype 7,
[26:22] SPEAKER_02: the 7 being the hot immune revved up one,
[26:28] SPEAKER_02: that you could
[26:29] SPEAKER_02: See what happened, what would happen.
[26:31] You could predict,
[26:31] and I'll put the graphs in the post,
[26:33] but these are like night and day.
[26:35] I mean,
[26:35] this is just,
[26:36] you know,
[26:36] really a striking difference knowing these ecotypes.
[26:40] SPEAKER_02: So you've gone on.
[26:43] SPEAKER_02: It took a while to get this paper in Nature,
[26:47] SPEAKER_02: but while you've been working,
[26:48] SPEAKER_02: of course,
[26:49] SPEAKER_02: long since then,
[26:51] SPEAKER_02: and that was just a couple months ago when it was published,
[26:54] SPEAKER_02: but you've
[26:55] SPEAKER_02: to Beyond Melanoma to confirm this, right?
[27:00] SPEAKER_00: Yes, exactly.
[27:01] SPEAKER_00: So we've been,
[27:02] SPEAKER_00: you know,
[27:02] SPEAKER_00: since doing the work on melanoma,
[27:04] SPEAKER_00: you know,
[27:04] we've also been looking at patients with non-small cell lung
[27:07] cancer that have
[27:08] SPEAKER_00: received chemo immunotherapy.
[27:11] SPEAKER_00: And we're seeing,
[27:11] SPEAKER_00: you know,
[27:12] essentially a recapitulation of our findings from the
[27:15] tissue-based assays in the
[27:16] SPEAKER_00: paper,
[27:16] SPEAKER_00: but now in the blood in non-small cell lung cancer.
[27:19] SPEAKER_00: We're starting to look at muscle-invasive bladder cancer that are given a
[27:23] SPEAKER_00: combination immunotherapy with an antibody drug conjugate.
[27:27] SPEAKER_00: Um,
[27:28] We're also looking at mesothelioma or getting into therapies
[27:31] beyond the immune
[27:32] SPEAKER_00: checkpoint inhibitors like CAR T-cell therapy and multiple myeloma and,
[27:38] SPEAKER_00: you know,
[27:39] bispecific or the bite-based antibodies that are approaches
[27:44] that have a T-cell
[27:45] SPEAKER_00: engager and also target,
[27:46] SPEAKER_00: you know,
[27:46] SPEAKER_00: a
[27:47] SPEAKER_00: A molecule on the surface of the cancer cell.
[27:49] So we believe that different spatial ecotypes will be
[27:53] differentially associated
[27:55] SPEAKER_00: with response to therapy depending upon the mechanism of action.
[27:58] SPEAKER_00: For example,
[27:59] a large molecule like the antibody drug conjugate may need more
[28:03] vascularization
[28:04] SPEAKER_00: around the tumor in order to get there.
[28:06] SPEAKER_00: And there are
[28:07] SPEAKER_00: Spatial ecotypes that are associated with blood vessel growth.
[28:12] SPEAKER_00: And so we're starting to see some early promising data toward that end.
[28:17] And by looking at these major hallmarks of the tumor
[28:21] microenvironment biology,
[28:23] we believe that we can kind of help to fill this missing
[28:26] dimension in the clinic
[28:28] SPEAKER_00: for patients.
[28:29] SPEAKER_02: Yeah,
[28:30] so I do want to be clear here because as I just reviewed and
[28:35] you mentioned,
[28:36] SPEAKER_02: thank you,
[28:36] SPEAKER_02: Aaron,
[28:36] SPEAKER_02: about all the different,
[28:38] the pyramid of all these different immunotherapies that range
[28:41] up to neoantigen
[28:43] SPEAKER_02: vaccines,
[28:44] SPEAKER_02: but as you've already mentioned,
[28:46] SPEAKER_02: the bites,
[28:47] SPEAKER_02: the bispecific antibodies and,
[28:49] SPEAKER_02: you know,
[28:50] SPEAKER_02: the CAR-T or CAR-NK or CAR-Macrovate.
[28:53] SPEAKER_02: I mean,
[28:54] we got so many different choices for immunotherapy beyond
[28:57] checkpoint inhibitors and
[28:58] SPEAKER_02: new checkpoint inhibitors.
[29:01] SPEAKER_02: So we don't even know how to apply those.
[29:03] And this DM,
[29:04] the ability to tap into a blood test could help us in that
[29:09] regard.
[29:09] SPEAKER_02: Now,
[29:10] SPEAKER_02: today,
[29:11] SPEAKER_02: we have really crude ways of assessing whether a person might be good for
[29:17] SPEAKER_02: immunotherapy.
[29:19] SPEAKER_02: We've got this
[29:21] SPEAKER_02: The tumor mutation burden of the number of mutations,
[29:24] SPEAKER_02: the PD-L1,
[29:25] SPEAKER_02: which is the pathway of the T cells being braked,
[29:30] SPEAKER_02: and maybe the tumor DNA burden,
[29:33] SPEAKER_02: how much volume of.
[29:35] SPEAKER_02: We don't have much to work with.
[29:36] They're all ice pick type assessments to pick about would the
[29:41] person benefit from immunotherapy.
[29:43] SPEAKER_02: You have a holistic thing.
[29:45] SPEAKER_02: It's like the city...
[29:47] SPEAKER_02: that you can't actually get into now.
[29:49] SPEAKER_02: And the liquid is like the river of the,
[29:52] SPEAKER_02: it's like wastewater,
[29:53] SPEAKER_02: you know,
[29:53] SPEAKER_02: to detect what's going on,
[29:56] SPEAKER_02: you know,
[29:56] SPEAKER_02: in the world.
[29:57] You're getting a holistic view that transcends all these ice
[30:02] pick singlets.
[30:04] SPEAKER_02: And it's dynamic.
[30:05] SPEAKER_02: You can check it, you know.
[30:07] SPEAKER_02: It's amazing.
[30:09] SPEAKER_02: It's just amazing.
[30:10] SPEAKER_00: Thank you.
[30:12] SPEAKER_00: So,
[30:12] SPEAKER_00: you know,
[30:13] SPEAKER_00: one way that we like to think about it,
[30:14] SPEAKER_00: just going back to analogies,
[30:15] SPEAKER_00: is,
[30:16] SPEAKER_00: you know,
[30:16] we mapped out these different spatial microenvironments or
[30:18] neighborhoods related
[30:20] SPEAKER_00: to,
[30:20] SPEAKER_00: let's say,
[30:21] SPEAKER_00: you know,
[30:21] SPEAKER_00: the map of the city,
[30:22] SPEAKER_00: you know,
[30:22] SPEAKER_00: the city hall,
[30:23] SPEAKER_00: the urban environment,
[30:25] SPEAKER_00: you know,
[30:25] SPEAKER_00: the suburbs,
[30:26] SPEAKER_00: the rural areas,
[30:27] SPEAKER_00: etc.,
[30:28] SPEAKER_00: But with the liquid assay,
[30:29] we can zoom all the way out up into space and now have a
[30:33] satellite view of not only
[30:36] SPEAKER_00: the map of an individual tumor,
[30:38] SPEAKER_00: but we can start holistically,
[30:39] SPEAKER_00: as you mentioned,
[30:40] SPEAKER_00: look at multiple different cities,
[30:42] SPEAKER_00: multiple different areas on the map.
[30:45] SPEAKER_00: Eric Topol erictopol
[31:01] SPEAKER_00: We've been flying blind,
[31:02] SPEAKER_00: as I mentioned earlier,
[31:03] in the clinic in terms of the tumor microenvironment and its
[31:06] dynamics and how it
[31:07] associates with response to therapy and how the therapy
[31:10] remodels the tumor
[31:11] SPEAKER_00: microenvironment.
[31:12] SPEAKER_00: So we're now starting to look at,
[31:14] SPEAKER_00: you know,
[31:15] SPEAKER_00: serially,
[31:16] SPEAKER_00: for example,
[31:17] do these spatial ecotype levels change in a way that informs
[31:20] how patients will do
[31:22] SPEAKER_00: on a therapy?
[31:23] SPEAKER_00: We expect to see that this will likely be the case.
[31:26] We have some really early data,
[31:28] very exciting data that can show,
[31:29] for example,
[31:30] We were looking at a patient with metastatic melanoma that
[31:33] received different
[31:34] SPEAKER_00: regimens of immunotherapy.
[31:36] SPEAKER_00: And initially they showed progressive disease by imaging.
[31:39] The oncologists switched them from a single-agent immune
[31:42] checkpoint inhibitor to a
[31:43] SPEAKER_00: combination immune checkpoint inhibitor therapy,
[31:47] SPEAKER_00: probably with the intent to prime their immune system to respond.
[31:51] SPEAKER_00: And imaging still showed progressive disease.
[31:54] SPEAKER_00: So then they decided to switch them back to single-agent immunotherapy,
[31:58] SPEAKER_00: and they still showed progressive disease until eventually,
[32:00] SPEAKER_00: lo and behold,
[32:01] SPEAKER_00: they responded.
[32:02] And there was no way to tell from the imaging data that this
[32:05] was going to happen.
[32:06] SPEAKER_00: When we looked at the spatial ecotypes in the same patient over time,
[32:10] SPEAKER_00: we observed that the resistant ecotype SE4,
[32:14] SPEAKER_00: it rose initially while they had progression,
[32:16] SPEAKER_00: but then as soon as combination therapy was given,
[32:18] SPEAKER_00: it dramatically dropped down.
[32:21] SPEAKER_00: Whereas SE7, which is associated with benefit,
[32:24] SPEAKER_00: increased dramatically during combination therapy,
[32:27] SPEAKER_00: showing that there was priming actually happening.
[32:30] And so this suggests that the imaging result was something that
[32:34] we would call
[32:34] SPEAKER_00: pseudoprogression,
[32:35] SPEAKER_00: where there is inflammation in the tumor,
[32:37] SPEAKER_00: but you can't distinguish that from the tumor itself.
[32:41] And the spatial ecotype levels from the blood gave us insights
[32:44] that actually
[32:45] SPEAKER_00: anticipated the patient's ultimate outcome many months before imaging.
[32:49] SPEAKER_02: Yeah, I know.
[32:50] SPEAKER_02: I mean, this is extraordinary.
[32:51] SPEAKER_02: It's like spatial biology.
[32:53] of cancer and the tumor microenvironment in the tube of blood
[32:57] that you basically
[32:58] SPEAKER_02: unravel with methylation,
[33:01] SPEAKER_02: which is derived from these two different AI
[33:05] SPEAKER_02: extensive work, the spatial ecotypes and liquid ecotypes.
[33:08] SPEAKER_02: I mean,
[33:09] SPEAKER_02: putting it all together,
[33:11] SPEAKER_02: you know,
[33:11] SPEAKER_02: people have wondered what's going to be the importance of spatial omics.
[33:16] SPEAKER_02: And,
[33:16] SPEAKER_02: you know,
[33:16] SPEAKER_02: we've already seen,
[33:17] SPEAKER_02: you know,
[33:17] SPEAKER_02: I've written about spatial medicine where,
[33:20] SPEAKER_02: you know,
[33:20] SPEAKER_02: using spatial omics in patients that had a toxic epidermal necrolysis,
[33:26] SPEAKER_02: their lives were saved by detecting the
[33:29] the pathway of JAK kinase to give a drug that repurposed
[33:34] several people who had
[33:35] this because of spatial omics that had what otherwise would
[33:38] have been a fatal
[33:39] SPEAKER_02: condition.
[33:40] SPEAKER_02: And here now, it's a much broader use case.
[33:45] SPEAKER_02: Now,
[33:47] SPEAKER_02: can we get this,
[33:49] SPEAKER_02: anybody who has cancer and wants to have assessment in the future,
[33:54] SPEAKER_02: what immunotherapy,
[33:55] SPEAKER_02: what combination,
[33:57] SPEAKER_02: how will they respond?
[33:59] SPEAKER_02: How are we gonna get this from this monumental work into practice?
[34:06] SPEAKER_00: Yeah,
[34:06] SPEAKER_00: so we've spun out a company called Liquid Cell DX that is working on
[34:10] commercializing this technology for patients and will be
[34:14] conducting a number of
[34:16] SPEAKER_00: different studies to,
[34:17] SPEAKER_00: you know,
[34:18] establish that gold standard clinical evidence so that we can
[34:21] make those decisions,
[34:23] SPEAKER_00: that we can inform clinicians to say,
[34:25] SPEAKER_00: you know,
[34:26] SPEAKER_00: based on the spatial ecotype profile at a given time,
[34:29] SPEAKER_00: or even the dynamics of spatial ecotyping over time,
[34:33] These are the therapies that are most likely to work or not
[34:36] work for a given patient.
[34:38] SPEAKER_00: So that is the goal.
[34:41] SPEAKER_00: We're also continuing to do work in our academic labs.
[34:44] SPEAKER_00: So myself,
[34:44] SPEAKER_00: along with a co-senior author of the paper,
[34:47] SPEAKER_00: Professor Otto Chudhury at the Mayo Clinic,
[34:50] SPEAKER_00: we are doing studies in all these different areas.
[34:53] I mentioned some of them earlier,
[34:54] looking at different cancer types,
[34:56] different
[34:56] SPEAKER_00: types of therapies,
[34:58] SPEAKER_00: looking at the benefit of serial or dynamic monitoring of the tumor
[35:02] SPEAKER_00: microenvironment.
[35:03] SPEAKER_00: But ultimately, the goal is to get this out to patients.
[35:05] SPEAKER_00: Hopefully within,
[35:06] SPEAKER_00: you know,
[35:07] SPEAKER_00: a couple of years,
[35:07] SPEAKER_00: a few years,
[35:09] SPEAKER_00: that is the dream,
[35:11] SPEAKER_00: is to get this out to patients as soon as we can.
[35:13] SPEAKER_02: So the AI, you know, with that part, you know, you would have...
[35:19] SPEAKER_02: as the analytic part.
[35:21] But what would you predict about the expense of being able to
[35:26] get TME insights from
[35:29] SPEAKER_02: a tube of blood in the future?
[35:30] Is this going to be,
[35:32] I mean,
[35:32] you know,
[35:33] Grail started out at whatever,
[35:34] $900.
[35:35] SPEAKER_02: I mean,
[35:36] SPEAKER_02: also using methylation,
[35:38] SPEAKER_02: of course,
[35:38] SPEAKER_02: that probably is coming down and they're going to scale.
[35:42] SPEAKER_02: And as many other liquid biopsy companies that are just zooming in on
[35:47] SPEAKER_02: The mutation story, not on this.
[35:49] SPEAKER_02: In fact, I never thought we would see this.
[35:52] SPEAKER_02: Okay,
[35:53] SPEAKER_02: what you've done here is something that is,
[35:56] SPEAKER_02: you have to use your imagination to think it would be even possible.
[35:59] SPEAKER_02: But could this be done pretty inexpensively?
[36:03] SPEAKER_00: Yeah,
[36:03] SPEAKER_00: I mean,
[36:04] it should be competitive with existing assays that are already
[36:07] in use today by
[36:08] SPEAKER_00: major liquid biopsy companies.
[36:10] SPEAKER_00: Natera,
[36:11] SPEAKER_00: Gardens,
[36:11] SPEAKER_00: and others are doing methylation sequencing already,
[36:14] SPEAKER_00: so there's already a path to translation here,
[36:18] and it's not a lot of additional expense on top of more
[36:22] traditional comprehensive
[36:24] genomic profiling sequencing approaches that are already in use
[36:28] for therapy
[36:28] SPEAKER_00: selection.
[36:29] SPEAKER_02: Would the Stanford spin-out?
[36:32] Would it join up with a garden or a Natera so it gives them
[36:36] another layer of valuable data?
[36:39] SPEAKER_02: Or do you think this should be a separate path?
[36:43] We're exploring a number of different possibilities for
[36:46] commercializing the
[36:47] SPEAKER_00: technology,
[36:47] SPEAKER_00: for getting it up to patients as soon as possible.
[36:50] SPEAKER_00: Yeah.
[36:51] SPEAKER_02: Okay, so we're going to start taking some questions.
[36:53] SPEAKER_02: Jennifer says,
[36:54] SPEAKER_02: can you speak to if this is similar tech of CancerGuard,
[36:59] SPEAKER_02: which I think,
[36:59] SPEAKER_02: is that the garden one,
[37:01] SPEAKER_02: CancerGuard,
[37:02] SPEAKER_02: or which one?
[37:03] SPEAKER_00: Or the Cologuard, potentially.
[37:04] SPEAKER_00: Cologuard is what she probably means, right, Cologuard.
[37:08] SPEAKER_00: So,
[37:08] SPEAKER_00: I mean,
[37:08] SPEAKER_00: this is,
[37:09] SPEAKER_00: again,
[37:09] looking at specific methylation sequencing sites using an
[37:14] AI-based approach to
[37:15] SPEAKER_00: determine the tumor microenvironment,
[37:17] SPEAKER_00: and it's not being done
[37:19] SPEAKER_00: Yeah, no, I mean...
[37:44] SPEAKER_02: This is a theme that we see with AI,
[37:47] SPEAKER_02: is that whatever it's being used,
[37:50] SPEAKER_02: you find there's other dimensions of rich embedded data.
[37:54] SPEAKER_02: And I think that's what you were able to do here in a staged approach,
[38:00] SPEAKER_02: really quite extraordinary.
[38:01] And then you had to bring in many top centers around the
[38:06] country to get all the
[38:07] SPEAKER_02: samples and to have the brain trust.
[38:10] SPEAKER_02: Could you comment about that?
[38:11] SPEAKER_02: You mentioned Mayo and Yale.
[38:12] SPEAKER_02: Were there other ones as well?
[38:14] SPEAKER_00: Yeah,
[38:15] SPEAKER_00: we also worked with Washington University in St.
[38:18] SPEAKER_00: Louis,
[38:19] SPEAKER_00: and we had an independent cohort of melanoma patients,
[38:22] SPEAKER_00: a small one in the paper,
[38:23] SPEAKER_00: but we've since worked with them to do a large blinded validation of our
[38:28] SPEAKER_00: melanoma-based immunotherapy prediction where they...
[38:33] SPEAKER_00: Since we sent the patient samples,
[38:34] SPEAKER_00: we applied the assay,
[38:35] but we were completely blinded to their outcomes and then
[38:39] unblinded at the end.
[38:40] And we were able to stratify patients that were pre-treatment
[38:46] as well as we were
[38:46] SPEAKER_00: able to do in the original paper using the exact same cut point.
[38:50] SPEAKER_00: And so that was actually presented at AECR earlier this year.
[38:55] SPEAKER_00: We have other ongoing collaborations with various medical centers,
[38:58] SPEAKER_00: including folks at MD Anderson,
[39:00] SPEAKER_00: which is a powerhouse.
[39:02] SPEAKER_00: Yes.
[39:03] SPEAKER_00: Yeah.
[39:04] SPEAKER_02: Wow.
[39:05] SPEAKER_02: Yeah.
[39:05] SPEAKER_02: So this, before we got going, I was asking Aaron, how long did this take?
[39:11] And you might want to just say this was the better part of a
[39:14] decade or so.
[39:17] SPEAKER_00: Yeah.
[39:17] SPEAKER_00: So we,
[39:18] SPEAKER_00: you know,
[39:18] SPEAKER_00: went down this path,
[39:19] SPEAKER_00: you know,
[39:20] SPEAKER_00: back to my postdoctoral days when we were starting to develop,
[39:23] SPEAKER_00: you know,
[39:24] In the early days of next generation sequencing and really
[39:28] leveraging that large
[39:29] SPEAKER_00: amount of data to try to understand what was happening in the tumor
[39:32] SPEAKER_00: microenvironment.
[39:33] SPEAKER_00: But at the time, it was blunt instrument.
[39:36] SPEAKER_00: Bulk sequencing is what we had.
[39:38] SPEAKER_00: We didn't have single cell gene expression information.
[39:41] SPEAKER_00: We didn't have their spatial locations.
[39:43] But we worked on ways to sort of deconvolute that really
[39:46] complex data with sort of
[39:49] SPEAKER_00: figuring out the composition of the smoothie,
[39:51] SPEAKER_00: so to speak.
[39:52] SPEAKER_00: And then over time,
[39:53] SPEAKER_00: you know,
[39:55] SPEAKER_00: we had a paper in Cell about half a decade ago now,
[39:58] SPEAKER_00: so in 2021,
[40:00] SPEAKER_00: that laid out the concept of a tumor ecotype.
[40:03] SPEAKER_00: But again, that was early days.
[40:04] SPEAKER_00: We didn't have the spatial information yet.
[40:06] SPEAKER_00: And there was no liquid biopsy element either.
[40:08] SPEAKER_00: So it's been a long haul.
[40:11] SPEAKER_02: Yeah, that's amazing.
[40:12] Now,
[40:12] Jennifer Gardella,
[40:13] I love that name for what we're talking about,
[40:15] Gardella.
[40:16] SPEAKER_02: She's asked also, can this be drawn from a vein in the periphery?
[40:21] SPEAKER_02: I know the answer, but I'll let you explain.
[40:25] SPEAKER_00: So, I mean, wherever one can draw blood, one can apply this test.
[40:29] SPEAKER_00: It is so young.
[40:31] SPEAKER_02: Yeah, I mean, you're basically taking advantage of the fact that
[40:36] This is a dynamic biologic process and these cells from the
[40:41] tumor microenvironment,
[40:43] their signal is being expressed or found in the blood sample
[40:51] and the cell-free
[40:53] SPEAKER_02: sample DNA to get that.
[40:55] SPEAKER_02: So it's just extraordinary.
[40:57] SPEAKER_02: Other questions from the crew here, we've got almost 400 on board
[41:03] SPEAKER_02: It's been a fascinating thing.
[41:04] SPEAKER_02: When I read this paper,
[41:05] SPEAKER_02: I mean,
[41:07] it just blew me away because I never thought we'd have a blood
[41:09] test for the tumor
[41:10] SPEAKER_02: microenvironment.
[41:11] SPEAKER_02: Now,
[41:11] SPEAKER_02: with these great centers besides Stanford and Mayo,
[41:16] SPEAKER_02: Yale and all these,
[41:17] SPEAKER_02: MD Anderson,
[41:18] SPEAKER_02: don't they want to have this offered for their patients today?
[41:22] SPEAKER_00: There's definitely a lot of,
[41:23] SPEAKER_00: you know,
[41:24] SPEAKER_00: motivation to get this assay out there as soon as possible.
[41:27] SPEAKER_00: So the answer is yes.
[41:30] SPEAKER_00: And I think, you know, we're doing what we can to make that happen.
[41:33] SPEAKER_00: But the,
[41:34] SPEAKER_00: you know,
[41:34] SPEAKER_00: commercialization of the technology is ultimately going to allow us to,
[41:37] SPEAKER_00: you know,
[41:38] SPEAKER_00: as academics do what we do,
[41:39] SPEAKER_00: which is focusing on,
[41:40] SPEAKER_00: you know,
[41:41] SPEAKER_00: the basic science and building up the evidence and new technologies.
[41:44] SPEAKER_00: And then,
[41:44] SPEAKER_00: you know,
[41:46] SPEAKER_00: on the commercial end,
[41:46] SPEAKER_00: you know,
[41:47] SPEAKER_00: we can try to get it out to as many people as we can as soon as we can.
[41:51] SPEAKER_02: Well,
[41:51] SPEAKER_02: I,
[41:51] SPEAKER_02: you know,
[41:52] SPEAKER_02: having just reviewed this field of immunotherapies,
[41:57] SPEAKER_02: which have proliferated exponentially over the years since they first,
[42:02] SPEAKER_02: you know,
[42:03] SPEAKER_02: showed their efficacy.
[42:04] Now we have,
[42:05] you know,
[42:05] like 10 different classes of them and even more to come.
[42:10] SPEAKER_02: You know, we don't know how to use them.
[42:12] SPEAKER_02: Which combinations?
[42:14] SPEAKER_02: And how are they working before patients deteriorate or,
[42:18] SPEAKER_02: on the other hand,
[42:20] SPEAKER_02: go into remission or cure?
[42:21] So this is coming about at just the right time in the mission
[42:25] to move tumors from
[42:27] SPEAKER_02: cold to hot and get that immune system to kick in,
[42:31] SPEAKER_02: to squash the tumor.
[42:34] Do you see this,
[42:35] besides immunotherapy,
[42:36] being helpful to guide other types of cancer treatments?
[42:41] SPEAKER_00: So that's a really great question.
[42:42] So my answer is that every cancer therapy involves some
[42:48] relationship between the
[42:48] SPEAKER_00: cancer cells and the tumor microenvironment.
[42:50] The tumor microenvironment is going to be important for lots of
[42:53] different reasons.
[42:55] SPEAKER_00: So one can imagine
[42:56] SPEAKER_00: And even new therapies like RAS inhibitors,
[42:59] SPEAKER_00: which are just mind-blowing in their results in pancreatic cancer,
[43:03] they were also going to involve interactions with the tumor
[43:06] microenvironment.
[43:08] SPEAKER_00: So the answer is absolutely yes.
[43:10] It's just a matter of doing the studies to establish the
[43:12] evidence to link these
[43:13] SPEAKER_00: different neighborhoods to respond to resistance to various therapies.
[43:19] SPEAKER_02: I mean,
[43:19] SPEAKER_02: you know,
[43:19] SPEAKER_02: angiogenesis for cancer,
[43:21] SPEAKER_02: anti-angiogenesis for cancer never really got as big,
[43:25] SPEAKER_02: obviously.
[43:26] Do you see any role that you could bring them back by guiding
[43:30] the ecotypes?
[43:32] Or is that probably not realistic because it's more the immune
[43:37] cells getting into
[43:37] SPEAKER_02: the tumor than it is the blood vessels?
[43:41] SPEAKER_00: I definitely think there's,
[43:42] SPEAKER_00: you know,
[43:42] a way for spatial ecotypes to be useful in understanding
[43:46] whether patients are
[43:47] SPEAKER_00: likely to respond to,
[43:49] SPEAKER_00: you know,
[43:51] SPEAKER_00: angiogenesis inhibitors,
[43:52] SPEAKER_00: anti-VEGF,
[43:53] SPEAKER_00: or also in combination with,
[43:55] SPEAKER_00: you know,
[43:55] SPEAKER_00: immunotherapies like anti-TD1.
[43:57] SPEAKER_02: It opens up new possibilities that we didn't have a way to get at, so...
[44:03] SPEAKER_02: Alright, so have I missed anything, Aaron?
[44:05] Because we went through a lot of stuff and we try to simplify
[44:08] it because if you try
[44:09] SPEAKER_02: to read the paper,
[44:10] SPEAKER_02: which I'll obviously post that link to it when we get this up,
[44:15] SPEAKER_02: it is very difficult.
[44:16] SPEAKER_02: And if you start reading that and the supplementary material,
[44:19] SPEAKER_02: that'll take several days to try to do that.
[44:23] But did I miss anything that's of practical importance that
[44:27] people should take away
[44:28] SPEAKER_02: from this story?
[44:31] SPEAKER_00: No, I don't think so.
[44:31] SPEAKER_00: This was,
[44:32] SPEAKER_00: you know,
[44:32] SPEAKER_00: very wide-ranging and covered all of the,
[44:35] SPEAKER_00: you know,
[44:35] SPEAKER_00: major focal points of the paper.
[44:38] I would maybe only say that we do believe that the concept of a
[44:42] spatial ecotype
[44:42] SPEAKER_00: extends beyond cancer.
[44:44] SPEAKER_00: And so there may be opportunities to also leverage this technology and,
[44:48] SPEAKER_00: you know,
[44:48] SPEAKER_00: inflammatory diseases and other,
[44:51] SPEAKER_00: you know,
[44:51] SPEAKER_00: other pathologies that do involve changes in cellular neighborhoods.
[44:57] SPEAKER_00: And, you know, we'll be looking into that as well.
[45:00] SPEAKER_02: Yeah, that would be great.
[45:01] SPEAKER_02: I mean,
[45:01] SPEAKER_02: this is not just a story for cancer,
[45:04] SPEAKER_02: although that's the obvious one,
[45:05] SPEAKER_02: but may certainly have other use cases.
[45:09] SPEAKER_02: Well,
[45:09] SPEAKER_02: in closing here,
[45:12] SPEAKER_02: unless there's other questions from the crew that are on board,
[45:16] SPEAKER_02: just want to congratulate you.
[45:17] SPEAKER_02: This is extraordinary work.
[45:20] SPEAKER_02: I, for one, think it's been underappreciated.
[45:24] I've seen some reactions to it,
[45:26] and obviously you've seen infinitely more,
[45:29] but
[45:32] People get scared where they see spatial ecotypes and deep
[45:38] learning and machine
[45:40] SPEAKER_02: learning and how this all comes together.
[45:42] But the bottom line is not to miss here is that we have a new
[45:46] window in the tumor
[45:47] SPEAKER_02: microenvironment that's getting more important every week.
[45:53] so to be able to get to that and to be able to do that on a
[45:57] serial basis for which
[45:58] is a great unmet need we have nothing else that we can get that
[46:02] holistic picture
[46:05] transcends the things that are done today to help guide
[46:08] patients so thank you for
[46:09] your multi-year effort your long journey we'll follow this
[46:15] course I know there'll
[46:16] SPEAKER_02: be subsequent many papers I hope that there's success for this
[46:21] SPEAKER_02: Stanford Spinout to get this into the clinic as soon as possible.
[46:27] SPEAKER_02: And thanks for everybody joining us today.
[46:30] SPEAKER_02: I know this was not the easiest topic to follow,
[46:32] SPEAKER_02: but hopefully with Aaron's brilliant use of metaphors,
[46:36] SPEAKER_02: we're able to transmit the value of this just amazing work.
[46:41] SPEAKER_02: So in closing, thanks everyone.
[46:44] SPEAKER_02: Thank you, Aaron.
[46:45] And we'll get this all posted on Ground Truths with the links
[46:49] and some pictures to
[46:50] help visualize what is a tumor microenvironment and what does
[46:54] spatial ecotypes tell
[46:55] SPEAKER_02: us about changing the fate of a patient with cancer.
[46:58] SPEAKER_02: Thanks a lot.
[47:00] SPEAKER_00: Thank you, Eric.
[47:01] SPEAKER_00: It was an honor and privilege to be here.
[47:03] SPEAKER_02: Really enjoyed it.
[47:04] SPEAKER_02: Bye-bye now.
[47:05] SPEAKER_02: Bye.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.