Virtual Screening

An AI powered drug discovery and development solution for biopharma R&D to rapidly screen millions of molecules for a target protein and identify promising ones for downstream assays


Evaluation process delays drug development

Schematic illustration of docking a small molecule ligand (green) to a protein target (black) forming a protein-ligand complex
Schematic Illustration of Docking (Illustration by Scigenis - Own work, CC BY-SA 4.0)

Virtual screening is an efficient tool used in early stages of drug development for the prioritization of compounds. Potential drug candidates need to be extensively screened for both their binding ability and functional properties.

At Aganitha, we have developed pipelines for Virtual Screening tailored for SMOLs and Antibodies. Continue to know more about the Small molecule virtual screening. To learn more about our solutions in Antibody virtual screening, click here.

Lack of high throughput infrastructure and scalability

Screening for promising candidates virtually has been around for decades. However, it is still not a major part of commercial pipelines due to lack of high throughput infrastructure and the inability to automatically scale up for high throughput computational requirements.

Inability to build custom solutions

Difficulty in building cross functional teams that understand diverse domains while having the relevant technology expertise which is necessary to develop custom solutions tailored for a specific purpose


Our Solution

An AI powered high throughput virtual screening solution aiding drug discovery and development

Aganitha’s solution combines the traditional molecular docking methods with modern AI/ML methods to predict the binding affinity

Molecular Docking Methods:

  • Ligand preparation
  • Protein preparation using the 3D structure of the protein. Homology modeling is used if the protein structure is not available. The solution also utilizes AlphaFold, a deep learning-based 3D protein structure prediction method.

Molecular Docking Methods Enhanced by AI/ML:

  • Computational platform for high throughput screening and cloud-based Kubernetes environment with schedulers such as SLURM for auto-scaling and workload management.
  • Modern AI/ML method with ECFPs (Extended Connectivity Fingerprints), and molecular graph convolutions to predict protein-ligand binding affinity – helping to screen millions of molecules for the target protein and shortlist appropriate molecules for downstream assays.
The virtual screening process starts with the protein structure either available from the Protein Data Bank or predicted using AI/ML and Homology modeling techniques. Using Aganitha’s generative models, a de novo molecular library is created that contains a list of potential drug molecules based on their predicted binding energies and conformations. The selected compounds are filtered using AI/ML and traditional methods, and the candidate compounds are further evaluated in downstream analysis such as MD simulations.
Virtual Screening Pipeline

Key components & strengths

Traditional molecular docking methods enhanced by AI/ML

Predict the protein-ligand binding affinity faster at scale by combining traditional docking methods with modern AI/ML methods

ECFP, Molecular Graph Convolutions and Deep learning

Leverages AlphaFold, a deep learning-based 3D protein structure prediction method along with homology modeling, when protein structure is not available

High throughput Screening, Schedulers and Auto Scaling

Cloud-based Kubernetes environment for auto-scaling, with schedulers such as SLURM for workload management resulting in high throughput

HPC Cluster Environment and Custom Tools

HPC cluster coupled with an infrastructure that can scale to screen millions of molecules in a short time, and custom tools for specific use cases

Accelerated hit to lead optimization as part of the drug discovery and development process

Faster querying of structure activity relationships

In estimating protein-ligand binding affinity for a protein family/sub-family

Quick and Cost-effective Screening

Of millions of molecules with AI/ML powered high throughput methods and high-performance computing (HPC)

Speeds up searching for potential drug molecule

And filtering out of unsuitable molecules and compounds leveraging computational chemistry techniques

Scaling of computing infrastructure

With modern infrastructure design techniques and cloud-based Kubernetes environment for auto-scaling

Discover our offerings across the biopharma value chain

Learn more about our Virtual Screening