Rishi Verma

Machine Learning practitioner in the AI x Bio space @arcinstitute | ex @illumina, @carnegiemellon, @iitroorkee

[Rishi Verma @i_m_rive]

This is my first ever post on X!

And I’d love to share my first ever preprinted work PIE - Perturbation is Everything!

Generalizing across unseen cellular and experimental conditions is challenging but essential to therapeutics discovery. Our latest model PIE addresses this challenge by simplifying the learning objective and leveraging curated biological knowledge from public databases. PIE generalizes across a combinatorial space of perturbation studies, unseen cell lines and unseen perturbations.

Check out the preprint at https://t.co/9oOfpcIZaX

Huge shoutout to the team at @arcinstitute for driving this project:@abhinadduri , @beabevi_ , @basak_ersln , @davey_burke , @genophoria , @yusufroohani .

[Arc Institute @arcinstitute]

Most perturbation models learn only from experiments. Our new model, PIE, also draws on decades of curated biological knowledge to predict which genes respond to a perturbation, even in biological contexts and for perturbations it has never seen.

[Rishi Verma @i_m_rive]

@florian_marty I agree. We’ve barely scratched the surface and there’s a long way to go. More data, both in volume and modalities, would help us build better models.

[Rishi Verma @i_m_rive]

This is my first ever post on X!

And I’d love to share my first ever preprinted work PIE - Perturbation is Everything!

Generalizing across unseen cellular and experimental conditions is challenging but essential to therapeutics discovery. Our latest model PIE addresses this challenge by simplifying the learning objective and leveraging curated biological knowledge from public databases. PIE generalizes across a combinatorial space of perturbation studies, unseen cell lines and unseen perturbations.

Check out the preprint at https://t.co/9oOfpcIZaX

Huge shoutout to the team at @arcinstitute for driving this project:@abhinadduri , @beabevi_ , @basak_ersln , @davey_burke , @genophoria , @yusufroohani .

[Arc Institute @arcinstitute]

Most perturbation models learn only from experiments. Our new model, PIE, also draws on decades of curated biological knowledge to predict which genes respond to a perturbation, even in biological contexts and for perturbations it has never seen.

[Rishi Verma @i_m_rive]

@ericcosulea In case you’re looking for inspiration

[Rishi Verma @i_m_rive]

Two days ago, I had 3 followers. Now I have 30. Comment below if you want to know how I grew 10x :P

[Rishi Verma @i_m_rive]

Two days ago, I had 3 followers. Now I have 30. Comment below if you want to know how I grew 10x :P

[Rishi Verma @i_m_rive]

@dr_alphalyrae Maybe not a CRISPR moment, but check out PIE.

[Arc Institute @arcinstitute]

Most perturbation models learn only from experiments. Our new model, PIE, also draws on decades of curated biological knowledge to predict which genes respond to a perturbation, even in biological contexts and for perturbations it has never seen.

[Rishi Verma @i_m_rive]

@BoWang87 Thank you Bo! And congratulations on your new publication in Nature Biotech!

[Tyler @rezoundous]

Opus 5.5 cured my depression

[Rishi Verma @i_m_rive]

And preferably Irodov so you can find your people

[Anusha @__acbraingenome]

carry a book wherever you go

[Rishi Verma @i_m_rive]

@trq212 @bcherny 🙌

[Rishi Verma @i_m_rive]

@davey_burke Thank you Dave for supporting the project!

[Rishi Verma @i_m_rive]

@abhinadduri Thanks a lot for all your support Abhi!

[Rishi Verma @i_m_rive]

@yusufroohani Thank you for all your guidance Yusuf!

[Yusuf Roohani @yusufroohani]

There are two major issues holding back perturbation models from generalizing across conditions: system representation and signal identification

[Dave Burke @davey_burke]

At Arc we have a compounding delivery cycle for our AI/bio models, with model candidates every 6 months. Sometimes we have spinoffs where we explore ideas and cherry pick ideas forward. PIE is one such example. It explores how far we can push using prior biological knowledge sources to make direct differential expression predictions. Useful result as we build towards more general cell simulators. Congrats @i_m_rive, @genophoria, @yusufroohani and team.

[Rishi Verma @i_m_rive]

@DoctorYev @arcinstitute Please feel free to try the model on the Virtual Cell Challenge. Picking which datasets to train on and how to sample single cell outputs would be interesting.

[Rishi Verma @i_m_rive]

@k_t_sangster It’s essential I think. Perturbation data is limited in scale and even signal. Prior knowledge augments the representation of the biological system.

[Rishi Verma @i_m_rive]

@adibvafa Check out PIE

[Arc Institute @arcinstitute]

Most perturbation models learn only from experiments. Our new model, PIE, also draws on decades of curated biological knowledge to predict which genes respond to a perturbation, even in biological contexts and for perturbations it has never seen.

[Rishi Verma @i_m_rive]

@nalidoust @arcinstitute Rhaister helped us learn so much about the datasets. Happy to chat more!

[Rishi Verma @i_m_rive]

@genophoria @arcinstitute @yusufroohani Thank you Hani for all your support!

[Hani Goodarzi @genophoria]

Check out PIE! Our latest perturbation generalization model from @arcinstitute. Congratulations to everyone, but especially @i_m_rive and @yusufroohani!

[Arc Institute @arcinstitute]

The code is on GitHub, with model checkpoints and data on Hugging Face.

Learn more about PIE from co-authors @i_m_rive and @genophoria: https://t.co/znAz30HKeo

Or access the preprint: https://t.co/GD7I44ERhQ