DeepAutoQSAR

予測型機械学習モデルの学習および適用に向けた、自動化され、拡張性のあるソリューション

Create high-performing machine learning models using state-of-the-art methods

DeepAutoQSAR is a machine learning (ML) solution that allows users to predict molecular properties based on chemical structure. The automated, supervised learning pipeline enables both novice and experienced users to train and inference best-in-class quantitative structure activity/property relationship (QSAR/QSPR) models.

主な機能

Streamline model building with fully automated workflows

Automatically compute descriptors and fingerprints, create models with multiple machine learning architectures, and evaluate model performance.

Customize models to your project with unique project-specific descriptors

Provide your own descriptors in CSV format to be used in addition to or instead of those generated by DeepAutoQSAR for a wide range of applications beyond small molecules, such as polymers, organic electronics, catalysis, and more.

Ensure model optimization using best practices

Employ QSAR/QSPR best practices to minimize the likelihood of overfitting or misrepresenting a model’s performance while ensuring maximum predictive model performance.

Understand the domain of applicability using model confidence estimates

DeepAutoQSAR provides uncertainty estimates alongside model predictions to help determine how much confidence should be placed on predictions generated for candidate molecules which may lie beyond the model’s training set.

Visualize and analyze results to gain further insights

Visualize color-coded atomic contributions towards target property facilitating ideation of novel chemistry. Visualize and analyze DeepAutoQSAR metrics reports and plots in Maestro to enable further experiments — quickly learn what model architectures are most effective and how models generalize on holdout sets.

Scalable training to support small or large datasets

Use classical ML methods like boosted trees on smaller datasets while also supporting the largest scale QSAR/QSPR models using graph neural networks and other modern deep learning approaches.

事例紹介とウェビナー

Schrödinger の技術が、現実の研究課題の解決にどのように活用されているかをご覧ください。

ライフサイエンスウェビナー

Leveraging machine learning applications combined with physics-based modeling for drug discovery

Machine learning strategies in drug discovery are becoming increasingly popular and can be used in various areas.

ウェビナーを視聴する

ライフサイエンス事例研究

固形がんの治療を目的とした、高選択性Wee1阻害剤の高精度かつ計算機支援による探索

ケーススタディを読む

ライフサイエンス事例研究

包括的なデジタルケミストリー戦略を用いた新規D-アミノ酸オキシダーゼ阻害剤のリード化合物設計

ケーススタディを読む

ライフサイエンスウェビナー

Trends in modern hit discovery: How your ultra-large screens can benefit from machine learning

While traditional structure-based virtual screening has been successful in finding diverse hits to advance projects there is significant room for improvement of hit rates, diversity of hit chemotypes, available IP space explored, and the potency of unoptimized hits.