Eric Sun

neuroimmunology, antibodies, and biowarfare

[Eric Sun @EricSun50]

fluid singularities are not unphysical

[heiner @HeinrichKuttler]

BTW if finite-time blowup happens it arguably means NS was the wrong model for physical reality. And in fact there are plausible modifications (e.g., a time delay in the nonlinearity, solved by Varnhorn 10 or so years ago).

[Eric Sun @EricSun50]

It would be so funny if mRNA cancer vaccines “worked” because the mRNA itself is so immunogenic (adjuvant-like)

Systemically stimulate the immune system and hot tumors clear, cold tumors become hot

[David Usharauli @3DiMMUNE]

First, a paper suggesting that random-sequence mRNA vaccines enhance anti-tumor effects, and now another claiming that random, nonspecific T cells are sufficient to eliminate tumors. So much for antigen specificity.

#science
#immunology

https://t.co/aPAGZQOm3K

[Eric Sun @EricSun50]

@BetterCallMedhi @MelindaBChu1 @sama AlphaFold is open? That’s a new one. Point to one of the scientific breakthroughs AF causally contributed to bc there aren’t many

Did DM ever release their training setup? Is AF3 even open weight? You fell for DM over OAI PR

[Eric Sun @EricSun50]

Even though biology is experiment rather than hypothesis bottlenecked, we may see people become less attached to ideas and willing to change their minds because hypothesis generation and literature review now takes hours instead of weeks

Helpful for breaking dogma

[Eric Sun @EricSun50]

Protein sequences were evolutionarily optimized under metabolic cost constraints, with WFYHM as expensive amino acids

Since biotech relaxes the energy constraint, we should search WFYHM-enriched sequences because evolution hasn't searched this distribution as much

[Eric Sun @EricSun50]

@SuproteemSarkar humans don't have probablistic coherence either

[Eric Sun @EricSun50]

@bayeslord I expect most humans have cortex patches > 5mm^2 that are "enslaved" by the rest of the brain.

Organoid computers are potentially evil but not uniquely evil compared to growing brains via factory farming (humans included)

[Eric Sun @EricSun50]

@flowersslop @Michael_Druggan Hoping we summoned angels instead of demons

[Peak @TSupporter100]

@zeta_globin The humble clozapine:

[Eric Sun @EricSun50]

So your infosec team can run defensive cybercampaigns 😉

https://t.co/FXtHlMoqlR

[hero thousandfaces @1thousandfaces_]

why would you do this. why on earth would you do this. I don’t mean to be a doomer but WHY

[Eric Sun @EricSun50]

@aaronmring @EganPeltan Got it, so the auto/polyreactivity difference is number of antigens an antibody recognizes. My understanding then is

autoantibody:polyreactive antibody::number of recognized targets

which feels similar to

peptide:protein::chain length

[Eric Sun @EricSun50]

@aaronmring @EganPeltan I’m probably misunderstanding something

It feels like autoantibodies ARE polyreactive. If we magically assay’d all antigens, including historical, in the microbiome, I would expect an autoantibody that hits a single human antigen to light up microbial antigens as well

[Eric Sun @EricSun50]

@Psychonaut1001 something us will probably kill us before MCD is symptomatic

[Eric Sun @EricSun50]

@aaronmring @EganPeltan IGHV4-34 is good at recognizing bacteria and viral glycans

maybe evolution is justifying V4-34 self-reactivity because it protects against infection? there's a chance to reduce self-reactivity during affinity maturation

https://t.co/vSRXQ7BRM9
https://t.co/1Tj9HhAYJC

[Eric Sun @EricSun50]

@aaronmring @EganPeltan auto vs polyreactive antibodies maybe a false dichotomy and we just haven’t figured out the intended target of the “autoantibodies” yet

e.g. some mechanism data shows anti-TSHR antibodies cross react with Y. enterocolitica / H. pylori which was the “intended” target

[Eric Sun @EricSun50]

2% of all PDB structures are for COVID-19

[Eric Sun @EricSun50]

which also explains why pharma noticed that drugs that self-aggregated are more immunogenic

because they formed a repeating pattern to the B cell

[Eric Sun @EricSun50]

a highly ordered 2D array of a endogenous protein that's tolerated on its own breaks B cell tolerance.

because evolutionarily, viruses present as 2D crystal arrays of protein

repeated molecular geometry is itself a pathogen signal !!

[Eric Sun @EricSun50]

@daphnesolves Step 0.5 is predicting yes/no does it bind
Step 1 is affinity prediction to within 1-2 OOMs which is still unsolved by both simulation and ML models

There are around 10 more steps to have a drug ready for the clinic

[Eric Sun @EricSun50]

Founders with a chip on their shoulder are easier to manipulate