sarahconstantin's Shortform

sarahconstantin

links 4/24/25: https://roamresearch.com/#/app/srcpublic/page/04-24-2025

https://en.m.wikipedia.org/wiki/Spasmodic_torticollis cervical dystonia: a striatal problem!
https://pubmed.ncbi.nlm.nih.gov/22234840/ the "geste antagoniste" or "sensory trick" -- many motor disorders benefit from touching or prompting the body (either manipulating oneself or having someone else do it) to "get unstuck" from the undesired posture, immobility, or involuntary movement. at the turn of the 20th century, it was thought that this meant the disorder was psychogenic, but no; motor problems with real, visible brain lesions respond to such "sensory tricks."
https://austinvernon.site/blog/rockweathering.html Austin Vernon does a deep dive of rock weathering for removing CO2 from the atmosphere.
Raman spectroscopy and its variants can be really powerful. Non-contact, label-free detection of biological molecules.
- surface-enhanced Raman can detect viral RNA & even distinguish (some) different viruses https://pmc.ncbi.nlm.nih.gov/articles/PMC10088700/
- live & dead cells have different Raman spectra: https://www.mdpi.com/1424-8220/7/8/1343
- an EBV viral membrane protein in infected cells is detectable with Raman spectroscopy https://www.nature.com/articles/s41598-024-56864-y
- https://pmc.ncbi.nlm.nih.gov/articles/PMC10316396/ CARS (a Raman variant involving lasers) can tell apart DNA, RNA, and some different kinds of proteins and lipids (bands can be determined by the type of bond that's prevalent in the molecule)
- https://chatgpt.com/share/680a46e6-fac8-800f-bd52-fc3a2c34e998 OpenAI o3 suggests that if you want to remotely & nondestructively detect the presence of a virus on a surface, your best bet is probably CARS. (I don't understand CARS well enough to sanity-check the analysis, but individual linked papers say pretty much what o3 claims they say)
https://asteriskmag.com/issues/09/abundance-at-home Jasmine Sun, Clara Collier, and Kelsey Piper talk about what the "Abundance" philosophy means for civic engagement.
- shouldn't you be able to get your neighborhood together to build a playground, instead of only using "community input" as a veto? "Abundance" can easily fall into idolizing Robert Moses-esque government megaprojects and being nostalgic for the days when the community couldn't stop them...but an important part of small-a abundance is the freedom to build things for ourselves.
case studies of big pharma companies have shown quantitative efficiency gains from AI:
- https://www.bioprocessonline.com/doc/ai-breakthroughs-revolutionizing-pharma-tech-ops-at-roche-0001 Roche used predictive apps for yield and CQAs (critical quality attributes, pharma manufacturing speak for KPIs) resulting in 5-10% yield improvements and up to 50% reduction in CQA writeoffs (=bad batches)
- https://www.sanofi.com/en/media-room/press-releases/2023/2023-06-13-12-00-00-2687072 Sanofi's AI programs allowed it to accelerate some research processes "from weeks to hours", and increased the speed of identifying the best lipid nanoparticle for an application from "months to days". they also used AI to identify 80% of low inventory positions in their supply chain.
- https://www.pfizer.com/sites/default/files/investors/financial_reports/annual_reports/2022/story/data-and-ai-are-helping-to-get-medicines-to-patients-faster/ Pfizer used AI to analyze patient data in Paxlovid trials 50% faster, and reduced manufacturing cycle time of one step in the process by 67%.
- https://www.theinformation.com/articles/ozempic-maker-says-ai-is-finally-reliable-enough-to-produce-sensitive-documents LLMs allowed Novo Nordisk to cut its team of clinical documentatation writers from 30 to 3, and finish tasks in minutes that once took weeks. nobody got fired, but they froze hiring for writers and gave them more to do.
I vaguely remember an example of a medicinal chemist using AI in his workflow and finding it sped up manual tasks, but I can't find it now and most of the examples I can find are of medicinal chemists being disappointed in LLMs:
- https://iwatobipen.wordpress.com/2023/04/11/can-llm-understand-chemistry-memo-journal-jcim/
- https://practicalcheminformatics.blogspot.com/2024/10/silly-things-large-language-models-do.html
https://www.research.va.gov/research_in_action/Diabetes-drug-from-Gila-monster-venom.cfm GLP-1 degrades fast in the human body; for GLP-1 inhibitor drugs to be a thing, they had to find a long-half-life version, which was discovered serendipitously in Gila monster venom in 1992.
- https://en.wikipedia.org/wiki/Exenatide#History John Eng, (student of Rosalyn Yalow! ) was intrigued by a lecture he saw where gila monster venom caused lots of inflammation in the pancreas & thought he'd identify the chemical agent behind the effect. Turned out to be the GLP-like hormone exendin-4. he licensed it to pharma in 1996 and it became the diabetes drug exenatide, approved 2005.
  - clearly, more had to be done before we could get modern GLP-1 inhibitors -- even longer half-life and better potency through molecular modifications.
https://www.science.org/content/blog-post/you-can-probably-smell-planet-here Derek Lowe, ever the skeptic, isn't saying it's not life. The planet K12-18 b is full of dimethyl sulfide and dimethyl disulfide, which produce a strong cabbage-y odor, and on Earth are pretty much only naturally produced by phytoplankton and algae.
- is an abiotic origin physically impossible? of course not. but this is definitely an Intriguing Thing.
- because these molecules would break down under light exposure, they must be being continually replenished by some process. that's also consistent with life.
what we know about which drugs work in the clinic:
- https://www.sciencedirect.com/science/article/abs/pii/S1471489203001188 gene knockout phenotype (in mice) correlates well with the success of drugs targeting that gene
- https://www.nature.com/articles/s41467-019-09407-3 drug side effects in an organ system occur more when there's genetic evidence linking the drug target to a disease in that organ system.
- https://www.nature.com/articles/s41586-024-07316-0 drugs with genetic evidence linking the target to the disease are 2.6x more likely to succeed in the clinic
- https://ascpt.onlinelibrary.wiley.com/doi/pdfdirect/10.1111/cts.12980 drugs with unknown targets or mechanisms of action fail more
- https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2565686 drugs are more likely to fail because of poor efficacy than because of safety problems. cancer drugs and drugs from small companies have a worse success rate.
- https://www.nature.com/articles/nrd.2017.194 it's common for drug programs to continue even after evidence has come in that the mechanism/action pair doesn't work (sunk cost fallacy?) and these drug programs are especially likely to fail.
Biopharma looks at very, very few drug targets. There are only 854 targets for all FDA-approved drugs -- compared to 10,000-20,000 distinct proteins in the human body. About 1857 proteins both have experimental evidence of a link to a disease and belong to a known druggable class.
- https://www.proteinatlas.org/humanproteome/tissue/druggable
- implication being: it is SUPER HIGH VALUE to open up more classes of proteins to being "druggable".
  - for example by getting drugs inside cells (now possible with nanoparticles!) so that cytosolic and nuclear proteins are reachable.
inventors of new methods in biology often have physical or computational science backgrounds:
- proximity labeling (localizing molecules within the cell): Kyle Roux
  - https://news.sanfordhealth.org/research/kyle-roux-lab-work/
- GCamp6 calcium imaging, which allows you to see single neurons firing in real time: Tsai-Wen Chen, Karel Svoboda, Loren Looger
- genome sequencing: George Church (biochemistry)
  - https://en.wikipedia.org/wiki/George_Church_(geneticist)
- CRISPR:
  - Emmanuelle Charpentier: https://en.wikipedia.org/wiki/Emmanuelle_Charpentier
  - Jennifer Doudna (biochemistry): https://en.wikipedia.org/wiki/Jennifer_Doudna
- single cell RNA sequencing: Aviv Regev (CS, computational biology):
  - https://microbiome.mit.edu/team/aviv-regev/
  - https://en.wikipedia.org/wiki/Aviv_Regev
- https://en.wikipedia.org/wiki/Edward_Boyden Ed Boyden, inventor of both optogenetics & expansion microscopy, started in EE/CS before getting his PhD in neuroscience
- PALM (super-resolution microscopy with fluorescence): Eric Betzig, physics
  - https://en.wikipedia.org/wiki/Eric_Betzig
cryo-EM is old (80s) but atomic resolution cryo-EM of single particles didn't come along till 2020 https://www.nature.com/articles/s41586-020-2829-0
https://www.nature.com/articles/s41573-022-00552-x that great Jack Scannell article on predictive validity and why you want it

2Viliam2y

Maybe because it is easier to agree on a binary question ("should this be allowed or banned?") than on an open-ended one ("something should be done -- but what specifically?"). Give people a binary choice, and there is a chance that enough of them will agree. Give them an open-ended question, and most people will come with their own proposals, unwilling to support anyone else's proposal (unless they are allowed to do large modifications, which other people will oppose). (Here an individualistic culture probably makes it worse, because coming with your own proposal is high-status.) I guess most people have this experience, so they don't even try to make proposals to the public. Instead, if possible, they act alone, or with a small group of friends. [...] I am often too pessimistic, but I would expect many people to say "no", for reasons including "no specific reason, it just sounds suspicious to me: why you? why now? is this perhaps some kind of scam?" or "I will only agree if you update your proposal to include <my pet peeve, completely unrelated to the project>", plus a few people saying "I don't give a fuck, so I will vote 'no' in principle (maybe try to bribe me if you want my 'yes')". However, there are two situations near me where people somehow succeeded to build something for the community, so I should probably try to learn the details. In one case, it is a community garden: area between two garages was surrounded by a fence, and how there are tables and chairs, and about once in a month someone organizes some activities for kids there. In another case, in place of a former shop, a community center was set up. I think the latter is just one person's activity who someone got grant money to rent the place (maybe also made a non-profit for that purpose) so I would still kinda classify that as a pro-social grant-supported unilateral action. No idea how the former may have succeeded. BTW, you seem impressed by George Church very much, because you linked his

sarahconstantin2y20

thanks, fixed

sarahconstantin's Shortform

8

sarahconstantin's Shortform

8