Client success story
LLM’s for Target Discovery
Business Value Delivered: Better Drug Targets Identified Faster
Summary: Enabling Faster Target Identification Through Customized AI Tools
A mid-size biotechnology company wanted to identify novel drug targets using proprietary molecular data and publicly available literature, but data volumes were large and the data was in many disparate files and databases. bPrescient used large language model (LLM) technology augmented with retrieval-augmented generation (RAG) techniques to build a web application that allowed the client’s scientists to research and identify targets more quickly.
Challenge: Large and Disparate Data To Mine For Targets
The client wanted to expand their therapeutic discovery efforts by adding new targets to their screening efforts, but identifying and validating targets was time-consuming and inefficient. The client had a large amount of data collected in databases and had access to large volumes of scientific literature but integrating and mining that data to connect targets to therapeutic opportunities was proving difficult. The client’s delay in identifying new targets impacted their ability to move into new therapeutic areas.
Solution: LLM With RAG Provides Efficient and Trusted Insights
After assessing the client’s collected data and desired public data sources, and assessing public LLM models trained on scientific literature, bPrescient selected a large base model and augmented it with query-based guardrails using RAG that limited hallucinations and provided support for the LLM conclusions. In addition, the LLM query results were connected to literature references and enhanced with explicit queries to underlying data stores. bPrescient implemented a web application, deployed in the client’s cloud environment, that allowed the client’s scientists to securely interact with the LLM, view query results and supporting data, and save queries for later use. The solution was compared to public models using published benchmarks to ensure additional value was being added.
Outcome: Targets Identified Quickly and Accurately
The LLM application allowed the client to more efficiently mine large amounts of disparate data from public and internal sources, and to gain insights into potential novel drug targets. The results their scientists gained from using the application spurred further research and allowed the client to move into new therapeutic areas.
Let’s talk
Tell us about your toughest R&D challenge and let’s discuss how bPrescient can solve it.
