Aditi Qamra (@Itti_Q) header
Aditi Qamra (@Itti_Q) profile picture

Computational biologist | Cancer researcher. Ph.D. @astar_gis Postdoc @UHN. Early drug development and biomarker discovery @RocheCanada. Now @arteraAI

Toronto, Ontario
Joined May 2012
Photo by Aditi Qamra (@Itti_Q)Photo by Aditi Qamra (@Itti_Q)
Time for the biggest drop of the week: auto-optimizations of eleven bioinformatics packages, focusing on read alignment. Each auto-optimization includes: a new unit test suite, the optimizations stacked on top, and a list of issues found during the validation.
9
39
8
161
17,083
It indeed feels like the entire "AI×bio" industry is under a massive illusion about the hurdles of drug development. I wonder how long people will continue raising billions on the basis of "we'll run AI on all scientific literature and multi-omics data, and we'll cure all disease in 10 years" This is really a great piece.
In certain tech circles, it has become low-status to suggest that AGI won't automatically solve everything. I think models will continue to get smarter, but intelligence is often not the main bottleneck to physical world change. More below 👇 writingruxandrabio.com/p/int… Some points the article makes: 1. Intelligence is often not the main bottleneck Increasingly, when I talk to people at AI labs, I keep hearing the same thing: that working on regulatory barriers to the diffusion of medical innovation is a waste of my time. AGI will solve all of that, they assure me, with the pitying look one might give a naive hamster awaiting its slaughter. I think they're wrong. We've had the technology to build better housing for decades, yet it's more expensive than ever: this is because housing is a question of political will, not just capability. Something similar goes for medicine. 2. This is just a microcosm of a broader problem: a dismisal of institutional and cultural barriers, as well as the mundane and the human. The regulatory barriers to AI diffusion in medicine is just one example of the kind of stuff that tends to be dismissed by the AGI-pilled people. I think in general, there is an underweighting of cultural and institutional shaped problems, as well as of the boring and mundane. Maybe this has to do with the "totalizing' feeling of being close to reaching the AGI-God: It becomes hard to care about other things. But to me, problems like a decline in reading or the slop-ification of our minds due to these technologies are equally important. They do not seem to be seen as equally serious. 3. This has become a monoculture. Certain SF tech circles (not everyone) have hardened into something close to a monoculture, where it's now genuinely low-status to suggest AGI might not be wholly transformative on its own. The people who were early and right about AI made fortunes. Being right about this one big thing has created a sort of "halo effect" where many think they're right about everything. To question the most extreme claims about the effects on the real world often means you will be marked as dim, low-IQ, insufficiently "AGI-pilled." This often results in self-censorship. Other very smart tech people privately acknowledge the bottlenecks to real-world diffusion, but do not really talk about it. Because the outside world has long been skeptical of AI (and continues to be), many in these circles have come to see themselves as dissenters. But I think AGI-maximalism is the safe and status-conferring view within their circles. 4. The real world has many barriers to progress -- more detail. The "real world" has many barriers to diffusion that those who think intelligence is all that matters seem to ignore. I remember a Progress Conference in 2023 where the economist Tyler Cowen made exactly this argument to a room full of AGI-pilled attendees, many of whom were frankly incredulous. The capabilities, he suggested, would arrive faster than the changes they were meant to produce; GDP growth would disappoint relative to the hype, and there would be no massive job displacement. So far, the data seem to bear him out. I can speak about my own area, which is medical progress, which is often bottlenecked by institutional and regulatory or policy barriers and also by the sheer difficulty of producing experimental biological data and the inherent limits of biology. One example, beyond my usual clinical trials points: our patent system rewards novel chemical matter much more than novel biology (finding new targets). This means whoever validates a new target bears the risk but finds it hard to capture the reward. The result is "target herding": a lot of drugs against the same de-risked pathways, meaning a lot of biological space goes underexplored. Some of best-funded AI-bio companies (Chai Discovery, Isomorphic Labs) target molecule optimization, so the tractable part, not the frontier of finding new targets. It turns out even AI companies respond to legal and economic incentives! Does this matter? I think so... We fixate on AGI gods, while more mundane problems pile up unattended. The social fabric is fraying in plain sight. More and more young adults cannot afford a house or sustain a relationship. People pour their youth into gambling apps, some of them now made more powerful by the very AI we keep discussing. People are de-skilling as they hand more thinking to machines. Every one of these problems needs people working on it. Being in the thrall of digital deities that will magically dissolve all our problems pulls attention away from the simpler, more human troubles that are already here. Note: I want to mention there was an earlier version of this X post that got a lot of reactions. I deleted it, because it became quite controversial and it was interfering with my work. One common rebuttal was that I am "making up a person to be mad at". I have modified my X post now to capture more of the nuance that I hope is found in the original article. For example, not everyone in SF thinks this and my new post makes that more clear. The X post is a summary though and I encourage people to read the whole article. This is also a way to start a conversation, not a definitive statement.
1
12
4
82
11,823
Everyone wants to cure all diseases but nobody wants to do no hard ass biology experiments
43
49
6
538
43,129
Aditi Qamra retweeted
Introns have a hidden regulatory role! 🧬🎉 Delighted to share our latest paper showing that inefficiently spliced introns and spliceosomal proteins direct RNA methylation, engaging RNAi to silence retrotransposons and regulate gene expression nature.com/articles/s41467-0…
🤖 Made with AI
8
85
2
274
31,382
To be clear, not saying sc is a bad tool. But 'predicting novel targets' means engaging with pathophysiology you're trying to reverse. Cell culture doesn't have pathophysiology. The models aren't built for novel targets in complex disease. Good work, for a different question.
2
17
1,611
Aditi Qamra retweeted
Our recommendations include: → Transition to genome benchmarking It is time to move beyond isolated small-variant metrics towards evaluating entire genomes against complete, diploid reference sequences to accurately evaluate and benchmark complex genetic variation.
1
1
92
Aditi Qamra retweeted
As an AI optimist, I do think AI will fundamentally transform biology and medicine. But after 25 years in academia and biotech and doing biology and quantitative biology at the highest levels, I think many AI companies encounter two very hard truths as tech companies have in the past. First, drug development is constrained by regulatory requirements, clinical trial frameworks, and their incentive structures that determine what can be discovered, tested, and ultimately brought to patients. Better models alone do not solve those bottlenecks. Second, AI is only as powerful as the questions it is asked. Biology is an extraordinarily difficult field because we still lack the right conceptual frameworks for understanding how living systems work. AI will accelerate discovery, but only if someone knows which problems matter, what data are informative, and how to distinguish mechanism from correlation. In my experience, people who can do that are incredibly rare. Many areas of biology have only a handful of scientists with both deep biological intuition and the mindset needed to fully exploit AI. They know which data should exist but don’t, which experiments will change our understanding, and which questions are worth asking in the first place. AI companies recruit elite computer scientists with extraordinary urgency. If they want to transform biology, they will need to recruit elite biologists with the same urgency. Otherwise, they’ll be limited by the quality of the biological questions those models are asked to solve.
I wish Anthropic well, especially since the history of drug development is a graveyard of tech companies trying to do drug discovery, mostly because tech companies just don’t get the complexity, uncertainty, revenue potential and timelines for biology. Anthropic does seem different since they’re much more science-oriented than the average tech company, so I hope it works.
23
34
3
188
26,809
Aditi Qamra retweeted
Claude Science is incredible. I gave it some sequencing data, and in 8 hours it did a full analysis, generated figures, wrote a paper, submitted it for publication, got rejected, revised and resubmitted, got rejected again, it is now applying for positions in industry
102
663
59
9,006
498,979
Aditi Qamra retweeted
Hope "AI Scientists" don't go the same trajectory as many bioinformatics tools. 10000 flavors. Impossible to know who is better for what purposes. Everyone benchmark maxxing. Users bewildered & confused. Hopefully they will all be equally useful rather than equally useless.
21
29
5
349
19,992
Aditi Qamra retweeted
Shows very nicely why the current “virtual cell” efforts are just a waste of compute. People throwing data at models hoping that they learn principles without the proper context, rather than putting effort in to actually capture the underlying biochemistry/biology.
I made a map of gene regulation as one integrated control system, from DNA → RNA → protein. 🧬 It contains 38 mechanisms grouped into 7 layers. It has a few recurring principles, such as: 🔓 accessibility 🏷️ reversible marks ⏱️ and kinetic coupling. 1/2
4
6
1
78
13,221
AI has transformed how we design therapeutics. But targeted delivery is still an expensive guessing game. Today @BobbyHollings and I are launching @deliverome with @beckypferdehirt and @radialscience at @AsteraInstitute, to fix that. 🧵
10
39
13
228
33,019
This is a great move in my view. Despite literally being called PolyCYSTIC ovary syndrome, PCOS does not actually involve cysts! The nomenclature stems from a misconception from 1933, when doctors observed many follicles on ultrasounds of women with PCOS & called them cysts.
6
9
2
120
10,150
Aditi Qamra retweeted
this
pretty crazy how five years ago vaccines were the most devastating thing you could ever do to your body but nowadays its fine to inject chinese peptide from Temu because you cant stop snacking after dinner
4
7
128
19,361
A scientist is helping uncover the role of the immune system in endometriosis—while managing the disease herself. Learn more on #WorldHealthDay: scim.ag/3ViJIUV @NewsfromScience
4
49
5
220
25,443
Aditi Qamra retweeted
New AI paper from us this week. When my student first showed me his initial findings, I really didn’t know what to make of them. I felt that this was an interesting but curious loophole phenomenon that would shortly be closed. I was very wrong. arxiv.org/abs/2603.21687
59
255
72
1,487
479,915
Excellent example of the core issue I was trying to get at wrt "virtual cells" - optimizing a prediction objective will not necessarily provide an accurate representation of physical or biological phenomena. It's extremely important to be very clear about this. 1/
Can an AI model predict perfectly and still have a terrible world model? What would that even mean? Our new ICML paper formalizes these questions One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
16
100
5
781
94,421
How is the start-codon selected for translation ? @iaincheeseman and colleagues suggest that cells globally control stringency of translation initiation. Low-efficiency initiation sites are preferentially repressed in mitosis, suggesting roles during the mammalian cell cycle.
2
35
2
175
22,837
Aditi Qamra retweeted
Women lose >90% of their #eggs by their mid-30s, with persistent #DNADamage driving this loss. But why isn’t this damage repaired? Our new work in @CurrentBiology, led by @ninadini_sharma, reveals the causes of high DNA damage in aged oocytes.🧬💥 bit.ly/48e35Ee (1/13)
13
251
24
1,257
208,598
Relationship between chromosomes (orange) and the mitotic spindle (green) in four stages of cell division, as seen by Bessel Beam plane structured illumination microscopy cell.com/fulltext/S0092-8674…
7
150
19
656
64,911
Aditi Qamra retweeted
Classifying cancer-associated fibroblasts—The good, the bad, and the target: Cancer Cell cell.com/cancer-cell/fulltex…
2
52
204
15,881