Reticular (YC F24)

Ending genetic disease, starting with embryo selection

[Reticular (YC F24) @ReticularAI]

Genomes are long but redundant. Not great for pretraining.

To fix that, we've developed a lossless tokenization strategy that shortens sentence length by >2000x.

For pretraining, we achieve a 0.55 validation CE on All of Us.

Public replication on 1KG linked below.

[Reticular (YC F24) @ReticularAI]

@jb_coger Ty!

[Reticular (YC F24) @ReticularAI]

Genomes are long but redundant. Not great for pretraining.

To fix that, we've developed a lossless tokenization strategy that shortens sentence length by >2000x.

For pretraining, we achieve a 0.55 validation CE on All of Us.

Public replication on 1KG linked below.

[Reticular (YC F24) @ReticularAI]

Public replication here: https://t.co/8WkNAgYR70

Scaling laws coming soon.

[Reticular (YC F24) @ReticularAI]

Infertility & pregnancy loss have lots of hidden costs. Not just financial too.

To help, we run a Facebook support group of 50k+ women going through IVF. Join here: https://t.co/m7LPZrPlwo

https://t.co/wrgDbK4l4l

[Reticular (YC F24) @ReticularAI]

This work builds on NeuroRisk from @herasight
and colleagues, and GPN-Star from @yun_s_song's group.

We extend the rare-variant representation and test its additional contribution in All of Us. Credit to both teams for the foundational research.

[Reticular (YC F24) @ReticularAI]

Couples with a family history of severe autism have been reaching out, hoping we could find a genetic cause to screen for.

But the genetic signal was always sparse.

Now our new work on rare variants & GPN-star has increased performance by over 50%.

Another step towards ending genetic disease.

[Reticular (YC F24) @ReticularAI]

Learn more: https://t.co/UyiMS83M06
Technical blogpost: https://t.co/RVE0cgY8ZA
Kits starting at $399: https://t.co/fk89zJSBTW

[Reticular (YC F24) @ReticularAI]

This work builds on NeuroRisk from @herasight and colleagues, and GPN-Star from @yun_s_song's group.
We extend the rare-variant representation and test its additional contribution in All of Us. Credit to both teams for the foundational research.
https://t.co/RVE0cgXBa2

[Reticular (YC F24) @ReticularAI]

Just published our state-of-the-art polygenic scoring results.

We're emphasizing our benchmarking methods, especially how we split.

https://t.co/0TmxFf68O5

[Reticular (YC F24) @ReticularAI]

For anyone going to ICLR, we'll be presenting our poster at the GEM, LMRL, SciFM, XAI4Science, and MLMP workshops. Stop by if you're curious about bio interp!

[Reticular (YC F24) @ReticularAI]

A First Step Towards Interpretable Protein Structure Prediction

With SAEFold, we enable mechanistic interpretability on ESMFold, a protein structure prediction model, for the first time.

Watch @NithinParsan demo a case study here w/ links for paper & open-source code 👇

[Reticular (YC F24) @ReticularAI]

Longer 🧵 breaking down our paper 👇
https://t.co/OfMxg7jQBX

[John Yang @johnyang100]

If you’ve ever

- thought AI protein folding is magical ✨
- wanted more than a pLDDT score 🔎
- or just think mech interp in bio is cool 🤓

then read the 🧵 👇 on our first paper towards interpretable protein structure prediction just accepted to workshops at ICLR

[John Yang @johnyang100]

@NithinParsan shoutout @liambai21 and @etowah0 for beating us to the punch on translating mech interp to protein language models. excited to see where these techniques will lead given the dense amt of coevolutionary + structure info learned by these models.

[Reticular (YC F24) @ReticularAI]

Incredibly excited to launch publicly as part of @ycombinator's Fall 2024 Batch. Send us a DM if you're in the interpretability space or working in biotech. @johnyang100 and @NithinParsan

[Y Combinator @ycombinator]

YC F24's @ReticularAI makes protein AI models controllable and interpretable to help steer protein design with limited biological data, reducing costly validation cycles.

https://t.co/vIesWS3rkh

Congrats on the launch, @NithinParsan and @johnyang100!