Active Learning Applications

Accelerate research with machine learning

Modern, physics-based computational methods offer quantitative accuracy for a broad range of important properties for drug discovery and materials design. These methods are compute-intensive and often intractable at the scale of today's ultra-large chemical libraries. This classic accuracy vs. speed tradeoff requires scientists to make difficult choices in selecting a computational strategy which typically involves subsetting the data or resorting to faster (but less accurate) methods. 

Active Learning technology from Schrödinger offers a solution to this problem. Active Learning workflows train a machine learning (ML) model on physics-based data, such as FEP+ predicted affinities or Glide docking scores, iteratively sampled from a full library using Schrödinger's deep-learning powered QSAR platform, DeepChem/AutoQSAR. These models can generate predictions for new molecules much more rapidly, and can identify the highest-scoring compounds in a large library at a fraction of the cost. We have shown that these hybrid models’ speed to be comparable to  a ligand-based method.

Accelerate research with machine learning using Schrödinger’s Active Learning applications

Benefits of Active Learning

Screen billions of compounds using state-of-the-art machine learning models in a fraction of the time.

Handle Large Libraries

Today’s ultra-large chemical libraries for hit discovery contain tens of millions of unique scaffolds.

Explore Diverse Chemical Space

For 0.1% the cost, recover ~70% of the same top-scoring hits from Active Learning Glide that would have been found from brute force docking ultra-large libraries with Glide.

Find Potent Hits

Explore tens of thousands to hundred of thousands of idea compounds with Active Learning FEP+ against multiple hypotheses simultaneously in lead optimization to quickly identify compounds that maintain or improve potency while achieving other design objectives.

Explore Multiple Hypotheses

Identify high performing compounds in a fraction of time with Active Learning compared to brute force calculations where fast machine learning is used to model more computationally expensive physics-based modeling.

Get Results Faster

Active Learning Calculator

Glide (Dock All Compounds)


Active Learning Glide


For Compounds

Enter in the numbers for your project (type in the box or use slider) to compare compute time and cost.

*Estimated customer compute costs only, based on $0.06 per CPU hour and $.35 per GPU hour. Recommended hardware for AL-Learning Glide.
License costs are not included. Contact us for a quote.

Drive research in a multitude of project areas

Schrödinger currently has multiple software applications that can utilize Active Learning to accelerate the discovery process.

The following is a list of what is currently offered:

Active Learning

Reduce hit discovery time and cost. Active Learning Glide combines the power of Glide, our industry-leading application for docking, with Schrödinger’s machine learning infrastructure to bring the exploration of ultra-large libraries within practical reach. Mine huge chemical spaces for hit compounds at a fraction of the cost, without sacrificing quality.

Active Learning

Reduce lead optimization time and cost by synthesizing fewer compounds to achieve design objectives. Rapidly explore vast regions of chemical space in lead optimization at lower cost while maintaining or improving protein-ligand binding affinity using accurate physics-based FEP+.


Hardware Requirements

Details of hardware and compute environment requirements to run Active Learning Glide.

High Performance Cloud Computing

Amplify the backend computing resources you need, when you need them, with the auto-scaling Virtual Clusters of Schrödinger in the Cloud.

Accelerating the Discovery of New Therapies to Fight COVID-19

As the biopharma industry races to develop treatments and vaccines for the global pandemic, it’s imperative that we both broaden our gaze and quicken our pace.

AI and Deep Learning in Drug Discovery and Materials Science

Schrödinger’s Machine Learning experts discuss AI and Deep Learning’s role in Drug Discovery and Materials Science now and in the future.

Discovery at Scale for Small-molecule Drug Discovery

Experts discuss the latest innovations in Small-molecule drug discovery at Schrödinger’s 2020 User Group Meeting.



Efficient Exploration of Chemical Space with Docking and Deep-Learning

Yang, Y.; Yao, K.; Repasky, M. P.; Leswing, K.; Abel, R.; Shoichet, B.; Jerome, S.

ChemRxiv, 2021, PreprintView
Reaction-Based Enumeration, Active Learning, and Free Energy Calculations To Rapidly Explore Synthetically Tractable Chemical Space and Optimize Potency of Cyclin-Dependent Kinase 2 Inhibitors

Konze, K.; Bos, P.; Dahlgren, M.; Leswing, K.; Tubert-Brohman, I.; Bortolato, A.; Robbason, B.; Abel, R.; Bhat, S.

ChemRxiv, 2019, PreprintView

Accelerate research with machine learning using Schrödinger’s Active Learning applications

CONTACT US to learn more

Back To Top