Diagnosing cancer by profiling the immune system | Hacker News Reader

Diagnosing cancer by profiling the immune system

github.com

219 pointsjostmey50 comments

This is really cool. Thanks for sharing. It validates the perspective of the late great Pieter Hintjens on this topic. https://github.com/hintjens/confessions/blob/master/ch08.txt

"In my body right now there is a holy war going on, and has been raging for years. My immune system has been doing its damned best to kill these rogue cells. And the rogue cells, unaware that they're destroying their own host, have been fighting back.

"The odds are on the cancer, of course, which is why this family of diseases is a major killer. Our bodies have to keep winning, year after year. Any given cancer has to win only once, and it's Game Over. The only way to beat cancer, really, is to die from something else first.

"Everyone fights cancer, all our lives long. From birth, our immune systems are hunting down and killing rogue cells.... We are all cancer survivors, until we're not. "

I discovered ZMQ after his death. Reading his last posts were perspective-shifting. I genuinely felt sad, both for him as an individual, and because many of his best works are not going to survive much longer than he did. Life is short and our impacts are transient.

I learned a lot from his writing. This view of the immune system really shifted my perceptions too. Maybe having your work help people, even years after your death, isn't a bad legacy.

I just skimmed his text, and it is apparent that he thought deeply about cancer, death, and living.

I don't think technology will eliminate all cancer, but from data I've seen, I'm convinced that cancer immunotherapy will help us control many cancers, and the technology is just getting started

He’s gained immortality through his work.

> Maybe having your work help people, even years after your death, isn't a bad legacy.

Isn't that, in some sense, the best possible legacy?

Wow, thank you for sharing that link. It was rational and humane and I found it very moving. I didn't know who he was, but I'm glad to know of him and his work now.

“I’m pretty sure, I’m not a doctor, but I’m pretty sure if you die, the cancer dies at the same time. That’s not a loss. That’s a draw."

- Norm Macdonald

Watching Norm Macdonald's standup in light of his then-undisclosed terminal cancer diagnosis is awe-inspiring.

I didn't even know he was sick..

He didn't tell anyone.

Indeed, we should be thankful there is no human equivalent of https://en.wikipedia.org/wiki/Devil_facial_tumour_disease

Thanks for this reminder about Pieter Hintjens. I have been greatly influenced by his writing, including his use of Alice & Bob, in "a protocol for dying" ! So witty, so clever - such a tragic loss.

Getting a readout of the immune system sounds like a huge advance. I'm a little surprised to first hear about it as a cancer diagnosis tool rather than as an autoimmune disease diagnosis tool

Trying to raise money for something like that! We have exciting preliminary results

I've been working on this area for quite long, and I think it's just a matter of time it becomes huge. Right now, a major roadblock is that single-cell RNA-seq, which may also give paired TCR alpha and beta chain data, is too expensive compared to bulk TCR-seq. The latter has limited resolution and more technical variability.

However, Microsoft Research has a preprint that employs bulk TCR-seq where they are able to detect Lyme, which is particularly difficult to do using antibody assays and often mistaken for multiple sclerosis: https://www.medrxiv.org/content/10.1101/2021.07.30.21261353v...

AFAIK, there are a few startups in stealth mode and non-stealth mode collecting big repertoire datasets. In those datasets, one can even measure the effects of vaccinations that occurred decades ago.

what's the approximate cost difference between scRNA-seq and bulk TCR-seq?

does the cost curve suggest that scRNA-seq costs will reach cost parity with bulk TCR-seq, or become more practical, within X years?

conversely, do you see a way for bulk TCR-seq to overcome issues with resolution and variability?

from the outside, it seems like sequencing is such a fundamental piece of understanding the human body yet technology seems incredibly limiting.

it's amazing what scientists have discovered so far. imagine what happens if we can help people observe cells more affordably and more accurately.

Bulk TCR sequencing cost around $500 to $1000 and captures under a million T cells. scRNA Seq costs around $10,000 and captures under ten thousand T cells. The price of sequencing is falling exponentially so my numbers can be out of date

These figures are approximately right. We've been able to bring the cost of single-cell RNA-seq + TCR-seq runs down to $3000 by doing targeted gene expression, which requires less reads. Besides, some newer microfluidics platforms are able to yield more cells, close to 2e4.

I would say that the major limiting factor is the microfluidics platform. Ideally, we would require something that has a yield within the same order of magnitude of bulk methods. To address this limitation, one can e.g. pre-sort activated cells. Otherwise, the number of receptors is typically not enough to be able to observe meaningful statistical patterns.

Parse biosciences and companies have developed methods to perform TCR sequencing without the microfluidics and it will be interesting to see how well it performs. These methods are all based on permeabilizing the cell and keeping it intact instead of encapsulating the in a droplet and then lysing it.

Microsoft Research was a partner in the joint collaboration. Pretty sure this was the research that culminated in the T-Detect Lyme product that Adaptive Biotech launched last year (definitely some Adaptive representation in the publication).

https://investors.adaptivebiotech.com/news-releases/news-rel...

I wonder if low-hanging fruit is just "immune system health". train on healthy vs unhealthy immune systems and give a status bar.

I've read that healthy individuals don't get cancer because their body nips it in the bud. Why couldn't we just figure out if people are pre-cancer (susceptible), like pre-diabetes?

Age is probably the biggest influence on a "healthy" immune system. I haven't kept up with recent research, but if memory serves the immune repertoire (diversity of T and B cell receptors) decreases essentially linearly with age.

Here's some supporting research from a quick search. https://journals.aai.org/jimmunol/article/192/6/2689/1552

Edit: some fun but possibly pontificating tangents

This seems to be part of the reason why vaccines have had relatively poorer efficacy in elderly populations. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3901832/ https://www.nature.com/articles/s41586-021-03739-1

I'm hopeful that cancer and immune profiling will be able to be used in tandem to (1) identify cancer early and (2) identify gaps in the immune repertoire to facilitate (3) design of antibodies that specifically target the cancer.

There's a huge variability in the age of thymic involution. Some people have a relatively active thymus at age 50 or 60, whereas in others it has switched off by age 25 or 30.

For those not familiar with immunology, the thymus is what populates your immune system with brand new T cells equipped with freshly selected T cell receptors.

This suggests interesting anti-aging opportunities, such as artificial thymi to repopulate your immune system and keep a healthy and diverse repertoire.

Induced Pluripotent Stem Cells are something oncologists have been looking at to create an unlimited number of T cell's, and PSC comes from the bones, so I wonder if bone health is an indicator of immune system response? Its quite common for people to break bones and the flat bones have the red bone marrow that makes the pluripotent stem cells. They can take Mesenchymal Stem Cells and convert them in the test tube into Induced Pluripotent Stem Cells to then become T Cells in the body, but I think they have as much control over the differention of T Cells once in the body like some oncologists using radioactive iodine for thyroid cancer only to find it dispersed around the body with some congregating in the thyroid and not having teh effect they hoped for.

In todays polluted world I wonder if we dont get enough omega 3's considering the role's they play and the triterpenoid's seem interesting.

The most surprising results I've seen is with a batch of histidine, but subsequent batches didnt solicit the same results which makes me wonder if the 1st batch was something like histidine dipeptides or some other reason like cellular pools being depleted of something. Considering histidine helps immune cells move through tissue to targets, aging suggests histidine levels are depleting even if an individual is healthy.

Same goes with glutathione, in disease and aging, it also depletes to dangerously low levels, problem is bacterial biofilms love it to reactivate themselves and until recently it was thought supplementing was a waste, but recent studies contradict this idea.

I wasn't aware of the variability in thymic involution. I guess most studies focus on averages. If you have links to more information about these outliers I'd be very interested in reading up on them.

The most recent literature that I'm seeing is showing the most pronounced time of thymus involution occurring around puberty, suggesting a programmed switch from a growing to reproductive phase of life.

"Regardless of the seemingly crucial role of the thymus in preserving homeostasis, its involution in humans and other mammals begins in childhood and peaks around puberty, resulting in an almost completely non-functional organ in aging." https://www.sciencedirect.com/science/article/pii/S156816372...

In this light the following makes sense (stopping or suppressing the switch to a reproductive life). "Castrating rodents before puberty or reducing the levels of sex hormones [e.g., by using Lupron, which desensitizes the luteinizing hormone-releasing hormone (LHRH) receptors] can attenuate or markedly recover the involution process in aging mice."

This study presents a gene expression marker for recent thymic emigrants and shows variability increases with age: https://insight.jci.org/articles/view/93739/figure/5

A much older study with different markers shows something similar: https://www.pnas.org/doi/epdf/10.1073/pnas.070061597

unrelated but spurred by this comment, have you seen studies exploring correlations between covid infection/severity and tonsils? a cursory google search yielded nothing.

See BRCA1 mutation

May I ask, what are your qualifications and how did you get a job in this field?

I'm surprised as well. I thought the problem with cancer is that the immune system doesn't react/detect it and therefore might not provide a discernible signal.

Sure, that's the problem with cancer. But to get there first, your immune repertoire has become less diverse and developed some holes due to infections, dysbiosis, aging, etc.

There's usually going to be some signal though even if it was just localized downregulation of some proteins on immune cells induced by a cancer to escape detection. The trouble there is that may be swamped in a systemic serum sample. Looking at subsets are more useful to improve signal to noise ratio like looking at extracellular vesicles released by tumour and immune cells for messaging purposes

Author here. We’re bringing this technology to market. We’ve identified specific use cases where these methods can impact clinical decision making. Here’s our LinkedIn page: https://www.linkedin.com/company/immunoscope . We’re looking to do our first raise

What studies, experiments, and/or trials has this gone through? Also, I see a list of "Publications" linked, but unsure how they relate to the work done here.

"Diagnosing cancer by profiling the immune system" is a bold statement to make while merely linking to a github repo. The readme still leaves A LOT of questions

The publications all appear to do what the README is describing, except for the part where the README says bugs are preventing some results from being reproduced.

Going through the stuff chronologically seems to help understand how the ideas got put together. The 2017 MS paper is the oldest of the four and mentions a weaker MSPrecise method using summary statistics. I mean… if we really are sure some two things are somehow related, you might as well throw ML at it and see what it comes up. It's just linear, no super fancy models.

I don't understand what Atchley factors are, here's the paper: https://www.pnas.org/doi/full/10.1073/pnas.0408677102

This was a plot twist in an episode of CSI:Miami several years ago.

Identical twins were implicated in a crime for which DNA was left behind, and they thought they were safe by being twins. But they also had distinct T cell (etc) signatures and that solved the case.

Due to mutations, identical twins probably don't even have identical DNA by the time they're born, let alone in their adult life. I don't think forensic DNA testing is a full sequencing though.

It is even more complicated than that! Some immune cells, like T cells, will actually edit their own DNA through a process named V(D)J recombination. So the immune system is actually editing some of its own DNA!

https://en.wikipedia.org/wiki/V(D)J_recombination

Like Lisp reader programming for dsl creation?

Right, but I mean even of non immune cells.

Human Chimeras[1] are even more weirder: one person with cells from two fraternal twins. Causes real problems with maternal[2] and paternal[3] DNA testing. Even more problems[4] if the twins are male and female.

[1] https://duckduckgo.com/?q=human+chimera

[2] https://wikipedia.org/wiki/Lydia_Fairchild — Fairchild stood accused of fraud by either claiming benefits for other people's children, or taking part in a surrogacy scam, and records of her prior births were put similarly in doubt . . .

[3] https://time.com/4091210/chimera-twins/ — the father had absorbed some of his twin's cells when he was a fetus, effectively becoming a chimera of himself and his brother. The man's previous child's DNA matched his . . .

[4] https://wikipedia.org/wiki/46,XX/46,XY Sex-chromosome discordant chimerism (XX/XY chimerism)

Those are interesting cases. A similar but more ubiquitous strangeness is somatic mosaicism, specifically in the brain where it seems that "no two neurons are genetically alike."

I was in line for the Terminator ride at Universal Studios when I stumbled onto this article a few years back. Couldn't stop thinking about it the rest of the trip.

https://www.scientificamerican.com/article/scientists-surpri... “The idea is something that 10 years ago would have been science fiction,” says biochemist James Eberwine of the University of Pennsylvania. “We were taught that every cell has the same DNA, but that’s not true.”

Edit:

Reading into this more again for the first time in a while. I'm amazed at how large some of these differences are. Many having 1 million base pair copy number variants.

"Single cell sequencing of endogenous human frontal cortex neurons revealed that 13%-41% of neurons have at least one megabase-scale de novo CNV, that deletions are twice as common as duplications, and that a subset of neurons have highly aberrant genomes marked by multiple alterations." https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3975283/

Couple this with...

"Megabase-scale copy number variants (CNVs) can have profound phenotypic consequences. Germline CNVs of this magnitude are associated with disease and experience negative selection. However, it is unknown whether organismal function requires that every cell maintain a balanced genome. It is possible that large somatic CNVs are tolerated or even positively selected." https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4772019/

I can't imagine our brains are just accumulating mutations of this size for no reason.

My favorite theory so far is that it has to do with memory indexing. https://www.frontiersin.org/articles/10.3389/fgene.2020.0037...

How do you get your data into this software though? What devices do you need for that?

As someone with a lot of auto immune diseases how do I get tested

Regarding cancer treatment, wonder how much more advancement did they get on that polio treatment. Where lot of test subjects with a brain tumor was treated by injecting polio virus into them.

https://www.uhhospitals.org/blog/articles/2020/01/clinical-t...

All the really interesting stuff is always written in python, (AI, science research stuff like this). Here I am and I haven't written any python in over 10 years now, ever since go overtook python after it's v3 debacle.

> go overtook python > All the really interesting stuff is always written in python

Those might not be super compatible statements ¯\_(ツ)_/¯

It could be in the venn diagram, "stuff rosywoozlechan has written in the past 10 years" and "really interesting stuff" do not intersect.

Reply on news.ycombinator.com