Sri Nandan Gondi

0 followers · 909 views

on the atlas — 41

highlights — 119

Our aim was to introduce as many mutations as we could in this 28 amino-region, including substitutions and insertions, the latter of which is a less common type of mutation in nature.
Diversifying gene therapy vectors with machine learning
Of those, approximately 110,000 produced viable viruses (many of our attempts were deep into the sequence space, where it is very hard to propose viable viruses). About 57,000 variants were farther than 12 mutations away from the AAV2 serotype. By generating more than two thousand sequences that were 25 or more mutations away, we decisively demonstrated the power of machine learning models to design diverse synthetic capsid sequences.
Diversifying gene therapy vectors with machine learning
After screening billions of potential sequences in-silico using machine learning models, we settled on ~200,000 designed variants which we experimentally tested for their viability.
Diversifying gene therapy vectors with machine learning
In this region, the average difference between two AAV serotypes is 12 amino-acids (often with few or no insertions).
Diversifying gene therapy vectors with machine learning
To test these methods, we focused on a representative region of the capsid
Diversifying gene therapy vectors with machine learning
Out of its two genes, the cap gene has a larger influence on the structure-function relationship. To start, cap expresses structural proteins VP1, VP2 and VP3, which interact to form the viral capsid from sixty copies of the VP monomers. Beyond contributing to the iconic icosahedral form, cap plays a critical role in determining tropism, the virus’ ability to infect a particular cell or organ type. Due to the capsid protein’s importance in capsid attachment to cellular receptors involved in tissue tropism, manipulation of the cap gene appears to be the best route to the selective tropism neede…
Introduction to AAV as a Gene Therapy Vector, Part 1
When converting a natural AAV into a gene therapy vector, its genome is dissected, manipulated, and reassembled to make room to include therapeutic genes (transgenes).
Introduction to AAV as a Gene Therapy Vector, Part 1
About 57,000 variants were farther than 12 mutations away from the AAV2 serotype. By generating more than two thousand sequences that were 25 or more mutations away, we decisively demonstrated the power of machine learning models to design diverse synthetic capsid sequences.
Diversifying gene therapy vectors with machine learning
After screening billions of potential sequences in-silico using machine learning models, we settled on ~200,000 designed variants which we experimentally tested for their viability. Of those, approximately 110,000 produced viable viruses
Diversifying gene therapy vectors with machine learning
When we started this study, it was unknown if machine learning models would be reliable for predicting the effects of mutations for variants beyond 5-10 edits to the original sequence. We expected this was possible, however, based on analyzing the diversity of sequences that have been isolated from natural sources. In this region, the average difference between two AAV serotypes is 12 amino-acids (often with few or no insertions). Nonetheless, we pushed the models to propose sequences with up to 29 substitutions and insertions.
Diversifying gene therapy vectors with machine learning
On the other hand, rational design approaches make targeted mutations to small sub-regions of the capsid based on hard-won biological knowledge, increasing efficiency at the cost of reducing scale.
Introduction to AAV as a Gene Therapy Vector, Part 2
The most promising route that is currently available uses AAV capsids, protein shells derived from the naturally benign human Adeno-associated Virus (AAV), as a vector to carry the healthy copy of the gene to the diseased organ or tissue.
Delivering gene therapy’s promise - Dyno Therapeutics
Instead of relying on our own fuzzy definitions of toxicity, we could perhaps instead defer it to a model capable of understanding phenotypes of toxicity more nuanced than ours could ever be.
A primer on why computational predictive toxicology is hard
One could imagine a world in which we have access to so much toxicity data that this problem ceases to matter — the model will figure it out.
A primer on why computational predictive toxicology is hard
The last part is important, because otherwise the existence of a subclass underneath the labeled class isn’t actually useful for a model to be aware of.
A primer on why computational predictive toxicology is hard
But the problem much more relevant to the toxicity discussion was the so-called hidden stratification problem; chest x-rays with a certain diagnosis label could be further subdivided into subtly different conditions with significantly different clinical outcomes.
A primer on why computational predictive toxicology is hard
But, while this goes far to include in more dense label information for each molecule, the underlying physiological impact of the toxicity is still missed!
A primer on why computational predictive toxicology is hard
But the biological relevance of many of these individual in-vitro assays to true organism toxicity is on shaky ground. One could say that any toxicity seen in-vitro will likely be seen in-vivo as well, but it’s unclear how true this is either.
A primer on why computational predictive toxicology is hard
When you enter this strange and unstable realm of conversation, you get a lot of information rapidly. I tend to find that almost everyone is captivating and loveable when I manage to talk like this. But when I do it with Johanna—especially in the first few years—it was like my entire mental landscape broke apart and all was possibility and flux.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
The type of person I’m assuming we’re looking for here is 1) someone that you will find fascinating to talk to after you’ve talked for 20,000 hours, 2) you feel comfortable with them talking through the hardest and most painful decisions you will face in your life, and 3) the conversation is wildly generative for both of you, in that it brings you out, helps you become. That is a very particular kind of conversation. You want to sample it as soon and as much as possible.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
That is perhaps the most solid dating advice I have, by the way—show the inside of your head in public, so people can see if they would like to live in there.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
They were obsessing over some obscure book, reading each other strange anecdotes from the 1800s settler communities in Norther Sweden, and she laughed like a hyena. I’d never been as intrigued by a human being before.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
Some people think Stein was lying when she said she wasn’t lesbian. And they are disappointed by that. (The fact is that most people Stein liked happened to be gendered like Alice.) But hers is the right attitude: you do not like a category. You like individuals. And you’re not born knowing which kind.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
thinking in categories would interfere with my ability to freely pattern-match for the particular type of individual I resonate with.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
That is: she is a singular human being, that I happen to love.
Looking for Alice - by Henrik Karlsson - Escaping Flatland
My pitch for working in biology is that you will be working to either prevent, or at the very least alleviate, the inevitable moment that Mother Nature decides to extract a pound of flesh from you, giggling and gnashing you between her teeth like a cat plays with a baby mouse.
Ask not why would you work in biology, but rather: why wouldn't you?
I can see the argument that it’s perhaps not good at designing binders for intrinsically disordered proteins or proteins that have never been experimentally characterized before. Still, for the set of targets that are experimentally well characterized, where you just want to generate interesting binders, I’m inclined to think that it’s transformative.
AI After Drug Development
Proteins rarely perform their biological functions in isolation but rather rely on protein–protein interactions (PPIs) to execute complex biological processes.
One-shot design of functional protein binders with BindCraft | Nature
As the team that worked on this project – none of us pregnant mothers – these results have not changed our eating habits significantly.
Report — PlasticList
Our findings suggest that the choice of container and time spent in plastic packaging could meaningfully impact the chemical content of takeout meals.
Report — PlasticList
The increase wasn't uniform across all chemical types: phthalate substitutes showed the biggest jump at 40%, while phthalates increased by a more modest 15%.
Report — PlasticList
food that spent 45 minutes in the takeout containers showed 34% higher levels of plastic chemicals overall compared to the same dishes tested directly from the restaurant.
Report — PlasticList
Should non-pregnant adults worry about this? We didn’t find strong enough evidence to conclude this with certainty. It’s probably not good for you to ingest exogenous hormonally active substances willy-nilly, but it’s also possible that you have bigger health concerns, like getting enough sleep, or exercising, or having purpose and meaning in your life.
Report — PlasticList
We do think there is enough evidence that plastic chemicals are bad for babies for this to be worthy of concern for parents, and further study by experts.
Report — PlasticList
The limits set by different agencies contradict each other, many of them haven’t been revised in decades despite advances in science, and real-world scenarios like chemical mixtures are understudied.
Report — PlasticList
However, we’ve emerged from this project with the view that current safety limits for plastic chemicals could be materially wrong.
Report — PlasticList
All in all, it seems likely that if the safe intake limits for these plastic chemicals were newly calculated today using modern science and data, they would be more consistent and lower, although it is possible most of them would still be above the levels that humans eat.
Report — PlasticList
That said, with the exceptions above, all of the foods we tested are safe to eat according to the FDA, EPA, and EFSA standards for chemical content in foods. So the question of plastic chemical safety in food comes down largely to whether you believe those organizations have set intake limits correctly.
Report — PlasticList
We were, like many others, asking ourselves if plastic chemicals would turn out to be the next public health crisis for humanity to overcome. We realized it was important to try and get closer to the true answer. Finding out how many plastic chemicals all of us really eat seemed like a good point to start, because (1) we could test that with precision and (2) if it turned out we don’t eat plastic chemicals, then maybe we shouldn’t care about their alleged health harms. So we got to work.
Report — PlasticList
i no longer feel like i’m being chased by a fire to make friends and force deep connections so i can fill some sort of gap in my life. i’m very fulfilled with the people i’m proud to love. i’ve never felt this way or this secure in my life. zero part of me feels fussed to make friends out of urgency; now it’s truly just out of my desire to get to know a cool person and see where it goes from there.
unpacking affection 💗 - introspection ft. harsehaj
i don’t distract myself by trying to get busier. it’s unhealthy, and eventually backfires. instead, i do things that help reinstill gratitude for my life, so the things that were bringing my mood down feel small in the bigger picture i’ve painted. regardless of how you decide to brute-force a good mood, the most important piece is that you act on it with urgency. your happiness should be a real priority.
brute-forcing a good mood ☀️ - introspection ft. harsehaj
it’s funny. much of colonization took our culture from us unwillingly, and now, to be accepted by them, we’ll willingly give it away.
observations of internalized racism 🪷
One could imagine a world in which we have access to so much toxicity data that this problem ceases to matter — the model will figure it out. But, as it stands, ClinTox is composed of only 1478 molecules, Tox21 + ToxCast with 15,000~ molecule, and TOXRIC with 100k+ molecules (in total, many of which lack all labels) — a sizable number, but a far cry from NLP-level token sizes. Perhaps pushing dataset sizes up even more alleviates this problem, but it feels more likely that alternate directions should be explored.
A primer on why computational predictive toxicology is hard
The ClinTox dataset in MoleculeNet does attempt to touch on a more complex notion of toxicity via a label denoting whether an in-vivo clinical trial using a given drug found that it was toxic. But clinical toxicity here is boiled down to a 1/0, no notion of whether the drug displayed hepatotoxic, cardiotoxic, neurotoxic, or otherwise properties. Another similar dataset is TOXRIC, which annotates a wide range of molecules with in-vivo, in-vitro, and qualitative toxicity measurements, specifying whether drugs display acute toxicity, carcinogenetic properties, respiratory toxicity, and 12 other c…
A primer on why computational predictive toxicology is hard
One could say that any toxicity seen in-vitro will likely be seen in-vivo as well, but it’s unclear how true this is either.
A primer on why computational predictive toxicology is hard
But the biological relevance of many of these individual in-vitro assays to true organism toxicity is on shaky ground
A primer on why computational predictive toxicology is hard
There’s a more fundamental problem here: the datasets we use to train predictive toxicology models are potentially too simplified for us to benefit from, even if models using them have perfect accuracy.
A primer on why computational predictive toxicology is hard
And this is one of the major problems with applying predicting toxicology at all — defining what is and isn’t toxic is hard! One may assume the FDA has clear stances on all these, but even they approach it on a ‘vibe-based’ perspective. They simply collate the data from in-vitro studies, animal studies, and human clinical trials, and arrive to an approval/no-approval conclusion that is, very often, at odds with some portion of the medical community.
A primer on why computational predictive toxicology is hard
The relationship between dose and toxicity is not always linear, and can vary depending on the route of exposure, the duration of exposure, and individual susceptibility factors. A dose that causes no adverse effects when consumed orally might be highly toxic if inhaled or injected. And a dose that is well-tolerated with acute exposure might cause serious harm over longer periods of chronic exposure.
A primer on why computational predictive toxicology is hard
While there are terms such as LD50, LC50, EC50, and IC50, used to explain the degree by which something is toxic, they are an immense oversimplification.
A primer on why computational predictive toxicology is hard