Manas Mahale

Rafa has come out swinging this morning 😂

GitHub - Manas02/Tanki: Terminal Based SRS Terminal Based SRS. Contribute to Manas02/Tanki development by creating an account on GitHub.

I made a Terminal based Anki clone (Spaced Repetition System TUI) for no reason !

github.com/Manas02/Tanki

It’s surprising how much you can do without JS

manasmahale.in/resume.html

This is what happens when you let computer scientists design molecules 🤯 Oral presentation @neuripsconf.bsky.social #NeurIPS2024 #chemsky

I am speaking at CPHI Middle East in Riyadh this week, on

"Artificial Intelligence Across Modalities - Small Molecules, Antibodies/Proteins, and Cell and Gene Therapy"

Please don't hesitate to get in touch if you are interested in AI & drug discovery! Cheers, Andreas

www.cphi.com/middle-east/...

PKSmart v3.0 is now live! pk-predictor.serve.scilifelab.se

PKSmart is an open-source app we built to predict human and animal pharmacokinetic parameters. Publication is coming soon and is also available for local implementation via pip install pksmart.

pksmart PKSmart: An Open-Source Computational Model to Predict in vivo Pharmacokinetics of Small Molecules

pip install pksmart

pypi.org/project/pksm...

Out-of-Distribution Validation for Bioactivity Prediction in Drug... Recent advances in machine learning for materials science have significantly improved the prediction of novel materials. Building on these methods, we have adapted them for drug discovery...

We translated a validation strategy in Materials to Small molecules.

A key problem in cross-validation is splitting of data. Here we propose a step-forward CV. People are used to scaffold splits, time splits etc. but these all have limitations.

Extended Activity Cliffs‐Driven Approaches on Data Splitting for the Study of Bioactivity Machine Learning Predictions The presence of Activity Cliffs (ACs) has been known to represent a challenge for QSAR modeling. With its high data dependency, Machine Learning QSAR models will be directly influenced by the activit...

About activity cliffs, sampling, similarity, and ML training:
onlinelibrary.wiley.com/doi/10.1002/...

✋

Great to see the communities re-forming!

Lately, on the side closer to cheminformatics, I've looked at the limits of predictive power of machine learning models

A cartoon robot using a chainsaw on a log

There’s a new Practical Cheminformatics post, “Some Thoughts on Dataset Splitting,” (with code and a robot cartoon) at practicalcheminformatics.blogspot.com/2024/11/some... .

related, this wonderful post by @greglandrum.bsky.social

greglandrum.github.io/rdkit-blog/p...

Yes, I began collecting various implementations of scaffold split, and there’s quite a variety of them. In general, your point that the scaffold split essentially resembles random sampling except in cases with clearly clustered data holds true when we examine those implementations.

GitHub - dehaenw/world-s-simplest-ECFP: the world's simplest ECFP the world's simplest ECFP. Contribute to dehaenw/world-s-simplest-ECFP development by creating an account on GitHub.

The world's simplest ECFP implementation: github.com/dehaenw/worl... #chemsky #cheminformatics

GitHub - Manas02/analogue-split: pip install analoguesplit pip install analoguesplit. Contribute to Manas02/analogue-split development by creating an account on GitHub.

Analogue Split creates train-test sets with a fraction (γ) of activity cliffs—molecular pairs with high similarity (above a threshold, ω) but with different activity labels. Results are visualized for metrics like accuracy, precision, recall, and F1-score across γ.

github.com/Manas02/anal...

GitHub - Manas02/analogue-split: pip install analoguesplit pip install analoguesplit. Contribute to Manas02/analogue-split development by creating an account on GitHub.

Analogue Split creates train-test sets with a fraction (γ) of activity cliffs—molecular pairs with high similarity (above a threshold, ω) but with different activity labels. Results are visualized for metrics like accuracy, precision, recall, and F1-score across γ.

github.com/Manas02/anal...

I just joined Bluesky as well... let's rebuild the same nice community in the cheminformatics/comp chem/bioinformatics/drug discovery etc. areas that 'the other place' was in the good old days 😊! Looking forward to reconnecting with people here, Andreas

@srijitseal.com

AndyGoesBsky.blogspot.com 😂

Thanks for this !