We've just released the first version of our Deep Learning Tuning Playbook! This is our attempt to distill our process for actually getting good results with deep learning. We emphasize hyperparameter tuning since it has been a large pain point. github.com/google-researc…

New blog post at Practical Cheminformatics, "Generative Molecular Design - We Need to Raise the Bar", practicalcheminformatics.blogspot.com/2023/02/genera…

Meet TxGNN, a model that utilizes geometric deep learning and human-centered AI to make zero-shot predictions of therapeutic use across a vast range of 17,080 diseases DBMI at Harvard Med Icahn School of Medicine at Mount Sinai Stanford University Harvard Data Science Initiative Harvard Medical School MIT CSAIL 1/9 txgnn.org

Online triplet contrastive learning enables efficient cliff awareness in molecular activity prediction ift.tt/pn0896K #chemrxiv_biochem

Based on the work of Derek van Tilborg, we developed Activity-Cliff-Awareness (ACA) loss in DL models for molecular activity prediction, we found that online contrastive learning enables efficient cliff awareness in the activity prediction. #QSAR #MachineLearning #DrugDiscovery

A collection of links to literature and code on generative artificial intelligence for molecular design and optimization github.com/AspirinCode/pa… #GenerativeAI #drugdiscovery #drugdesign

Introducing PINNACLE, a contextual graph AI model for comprehensive protein understanding PINNACLE dynamically adjusts its outputs based on molecular contexts in which it operates Providing outputs tailored to molecular contexts is essential for broader use of foundational

How well do your AI models perform on new molecular sequences? Yasha Ektefaie 🧵 👇 Understanding generalizability - how well an AI model works on new data - is crucial in biology. This challenge grows with foundation models, large pre-trained models that promise to better predict

UniTS: Building a Unified Time Series Model - Supports a universal task specification, accommodating classification, forecasting, etc - Remarkable zero/few-shot and prompt learning capabilities when evaluated on new data domains and tasks repo: github.com/mims-harvard/U… abs:

Excellent work!

I've added 3 new Active Learning tutorials to the Practical Cheminformatics Tutorial series, bringing the total number of tutorials to 24. github.com/PatWalters/pra…

SCIKIT-FINGERPRINTS Unlike in computer vision, in computational chemistry extracted low-level features are still competitive (if not SOTA) for many tasks. P: arxiv.org/abs/2407.13291 C: github.com/scikit-fingerp…

Excited to share our new paper on Contextual AI models for context-specific prediction in biology in @NatureMethods led by stellar Michelle M. Li (李敏蕊) rdcu.be/dOxQ7 Understanding how proteins work and developing new therapies requires knowing which cell types proteins act

Efficient Generation of Protein Pockets with PocketGen - The paper introduces PocketGen, a deep generative model designed for the simultaneous co-design of protein pocket sequences and their 3D structures, offering a significant advancement in drug discovery. - PocketGen

A good read on the perceived success of molecular machine learning. Don't forget your statistical tests guys! nature.com/articles/s4225…

AI models can trip on activity cliffs, where look-alike molecules behave in wildly different ways 🔀💊 Molecular representations often miss these cliffs, grouping structurally similar compounds together despite huge bioactivity differences Our new approach sharpens molecular

Excellent work by Wan Xiang Shen and the team! We’re very honored to further be adding the ACANet to the TDC Model Hub and have released the Molecular Property Cliff prediction task colab.research.google.com/drive/1kHdFG4g…

PocketGen: A groundbreaking deep generative model co-designs protein pocket sequences & 3D structures, achieving 95% success in generating high-affinity pockets. 10x faster than physics-based methods, it redefines drug discovery & enzyme engineering. Congratulations for such

📢 🧬 New preprint! Can we predict which cancer patients will benefit, before treatment begins? Wan Xiang Shen Immunotherapy saves lives but many patients don’t respond to treatment, and we still lack reliable tools to predict who will benefit We introduce COMPASS, foundation