Lightnews — Scholar-powered news

LightNews

Alex Rubinsteyn

@alexr.bsky.social

5.9K followers 1K following 1.2K posts

personalized cancer immunotherapy = genomics + immunology + machine learning + oncology (pirl.unc.edu)

Posts Media Videos Starter Packs

Alex Rubinsteyn @alexr.bsky.social · 13h

Enumerating permutations and then cross-referencing them against the literature. A huge boost vs doing the same thing yourself!

Alex Rubinsteyn @alexr.bsky.social · 13h

open.substack.com/pub/pirlblog...

Expanding the Unified Grammar for RNA Variants

A unified toolkit for identifying and integrating DNA and RNA variants. Part 2 of the Neoantigen Discovery with Long-read Sequencing series.

open.substack.com

Alex Rubinsteyn @alexr.bsky.social · 17h

It’s not certain and can’t be done entirely in the world of bits but a path to inverting Eroom’s law is coming into view, fun time to be alive!

Alex Rubinsteyn @alexr.bsky.social · 17h

There are still missing steps in the world of atoms but some wins can come from unlocking scale on “platform” research reagents (eg mRNA too cheap to meter), some wins from flexible/personalizable therapeutic platforms (like neoAg vaccines but more general and actually effective)

Alex Rubinsteyn @alexr.bsky.social · 17h

I’ve been skeptical about AGI = rapid medical progress but starting to believe a weak form of this

(Researchers assisted by the better AIs can move much faster than before, discover far flung areas niches they wouldn’t have considered & get good initial params for experimental design)

Alex Rubinsteyn @alexr.bsky.social · 2d

Which one is your favorite?

(1) Dishevelled
(2) Frizzled
(3) Pikachurin
(4) Sonic hedgehog

(other, post below)

Alex Rubinsteyn @alexr.bsky.social · 12d

Thanks!

Alex Rubinsteyn @alexr.bsky.social · 12d

@tszzl.bsky.social on bsky?

Alex Rubinsteyn @alexr.bsky.social · 12d

Link to feeds and accounts please

Alex Rubinsteyn @alexr.bsky.social · 13d

The quality gap between ChatGPT5 Pro / Deep Research and R1 is just too big for anything open or local to be likely useful (sadly!)

Alex Rubinsteyn @alexr.bsky.social · 13d

Haven't tried consensus -- is open?

Alex Rubinsteyn @alexr.bsky.social · 13d

Tried FutureHouse tools ~2 months ago and found them still to be solidly in the "botslop" category

Walls of text that are superficially reasonable looking but don't help you learn anything (and sometimes actually fall apart logically upon inspection).

Alex Rubinsteyn @alexr.bsky.social · 13d

ChatGPT Pro & OpenAI's Deep Research are extremely useful for me wrt lit searches and do a decent but imperfect job at summarizing the salient themes in a pile of papers (the models are a bit too gullible and sometimes get obsessed with a niche idea from a random preprint). But it's still a huge win

Alex Rubinsteyn @alexr.bsky.social · 13d

I've been stuck in a bio knowledge containment field.

I generally have used ChatGPT more for science since Pro/Deep Research surface better papers and seem to summarize them more skeptically and faithfully.

But surprised that I can't get Claude to even answer simple things now

Alex Rubinsteyn @alexr.bsky.social · 13d

Tried going back to Claude today and it’s…totally useless for biomed? Got this for many different questions in a row but can eventually get an answer if I write a long shpiel about why it’s really safe to tell me about an IHC CRO.

Alex Rubinsteyn @alexr.bsky.social · 13d

When mRNA synthesis too cheap to meter?

Alex Rubinsteyn @alexr.bsky.social · 14d

tl;dr MS guide

SRM: select specific precursor m/z, collide, measure specific fragment m/z

PRM: select specific precursor, collide, record all fragments

DDA: pass all precursors, measure m/z (MS1), select top N, PRM

DIA: select evenly spaced windows of precursor m/z, PRM

Alex Rubinsteyn @alexr.bsky.social · 15d

Meta-analysis of observational studies is such a cursed genre of paper.

eg Ahlqvist 2024 does a careful sibling analysis to show that apparent slight Tylenol:ASD risk is artifactual

Baccarelli 2025 throws both the full confounded cohort and smaller sibling study in the blender

Alex Rubinsteyn @alexr.bsky.social · 15d

We (me and @benjamingvincent.bsky.social) naively assumed that heavy isotope labeled peptide synthesis for targeted MS was a solved problem. Wish I had read more before ordering so many contaminated peptides:

pubmed.ncbi.nlm.nih.gov/35119480/

Light contamination in stable isotope-labelled internal peptide standards is frequent and a potential source of false discovery and quantitation error in proteomics - PubMed

In mass spectrometry-based proteomics, heavy internal standards are used to validate target peptide detections and to calibrate peptide quantitation. Here, we report light contamination present in hea...

pubmed.ncbi.nlm.nih.gov

Alex Rubinsteyn @alexr.bsky.social · 15d

I wonder if the political “big moral circle” of us over-educated folks is compensating for the actuality of day to day white collar culture.

Like, people want a nice lifestyles for their smallish families, take little risk, & often seem to have loose bonds to a greater community

Alex Rubinsteyn @alexr.bsky.social · 16d

Has anyone done MS peptidomics on FFPE samples?

If so, how did they get it to work? (imagine PRM gets thrown off by modified residues and DDA acquires too many bits of junk to make sense of original sequences)

Alex Rubinsteyn @alexr.bsky.social · 17d

Which New Year did early Israelites (ca ~1000BC) observe first: Spring or Fall?

Alex Rubinsteyn @alexr.bsky.social · 17d

Vaccine studies like to talk about their novel designs, cell therapies like to talk about their clinical responses

Alex Rubinsteyn @alexr.bsky.social · 17d

Listening to the newest TWiV with the heads of the European Society for Clinical Virology and ECDC, thinking “you’re the A team now”

Alex Rubinsteyn @alexr.bsky.social · 18d

But if you want to effectively call the cops then you probably want a well staffed and funded police department…

(eg Durham has very wimpy response times to even serious gun crimes and so far many different calls to the police haven’t resulted in any action on the individual this thread is about)