Idea
A protein-fragment encoder and generative platform enabling efficient fragment identification and design for drug discovery researchers.
Research Paper
Core Innovation
This paper introduces LatentFrag, a protein-fragment encoder using contrastive learning to embed fragments and protein surfaces in a shared latent space. It enables chemically realistic fragment generation conditioned on protein surfaces, improving fragment recovery rates and reducing computational cost compared to traditional virtual screening. This approach advances fragment identification and supports full ligand design workflows.
Market Size (TAM)
$20–50B TAM for drug discovery platforms; $2–10B SAM from pharmaceutical and biotech companies. Driven by demand for faster drug candidate identification and cost reduction in early-stage drug design.
Potential Customers & Pain Points
- Pharmaceutical Companies Needing Faster Fragment Hit Discovery
- Biotech Startups Developing Fragment-Based Drugs
- Computational Chemists Seeking Cost-Effective Virtual Screening
- Academic Labs Focused on Drug Design Innovation
Business Model
Subscription-based SaaS platform offering API access and custom integration services for pharmaceutical and biotech companies.
Competitive Landscape
- Schrödinger
- Cresset
- OpenEye Scientific
Implementation Challenges
- Integration with existing drug discovery pipelines
- Validation on diverse protein targets
- Adoption by traditional medicinal chemists
Validation Strategy
- Benchmark fragment recovery against standard virtual screening methods
- Pilot studies with pharma partners on real drug targets
- Demonstrate cost and time savings in fragment hit discovery
Research Paper Overview
Flow-Based Fragment Identification via Binding Site-Specific Latent Representations
Summary
Fragment-based drug design is a promising strategy leveraging the binding of small chemical moieties that can efficiently guide drug discovery. The initial step of fragment identification remains challenging, as fragments often bind weakly and non-specifically. We developed a protein-fragment encoder that relies on a contrastive learning approach to map both molecular fragments and protein surfaces in a shared latent space. The encoder captures interaction-relevant features and allows to perform virtual screening as well as generative design with our new method LatentFrag. In LatentFrag, fragment embeddings and positions are generated conditioned on the protein surface while being chemically realistic by construction. Our expressive fragment and protein representations allow location of protein-fragment interaction sites with high sensitivity and we observe state-of-the-art fragment recovery rates when sampling from the learned distribution of latent fragment embeddings. Our generative method outperforms common methods such as virtual screening at a fraction of its computational cost providing a valuable starting point for fragment hit discovery. We further show the practical utility of LatentFrag and extend the workflow to full ligand design tasks. Together, these approaches contribute to advancing fragment identification and provide valuable tools for fragment-based drug discovery.