Receptor.AI introduces a 5-Level Compound Selectivity Platform designed to improve precision in drug discovery by targeting highly similar protein variants. The platform addresses the growing need for compounds with high selectivity for disease-related protein isoforms, helping to reduce off-target effects and support more tailored therapeutic approaches.

The platform consists of five sequential levels, each focused on refining compound selectivity predictions. Starting with a proteome-wide ranking, the process involves explicit screening against off-targets, AI-based prioritization of protein sequence differences, and extends to assessing protein structure differences on an atomic level. This stepwise approach ensures progressive accuracy and selectivity, facilitating the identification of ultra-selective small molecules for challenging targets, including those with no known ligands or ambiguous structures.

To assess the platform, Receptor.AI conducted testing using proteins from the JAK and FGFR families, known for their significance in a range of diseases. The platform adeptly handled these proteins, demonstrating its capability to discriminate among highly similar variants, highlighting its utility in developing highly selective compounds.

‍

Key to this process are:

- Level 1's proteome-wide ranking that initiates the selectivity prediction by screening ~100K compounds across ~9.3k proteins (Figure 1).

‍

- Level 2 focuses on explicit screening against defined off-target variants, filtering out non-selective compounds.

- Level 3 and 4 employ AI models to prioritize sequence and structural differences, respectively, fine-tuning the selectivity towards the target proteins (Figures 2 and 3 respectively).

‍

- Level 5 introduces an innovative approach for targeting membrane proteins related to cancer, exploiting the unique composition of cancer cell membranes to achieve additional selectivity (Figure 4).

‍

‍

Benchmarking results confirmed the platform’s robust performance across all five levels, with particular effectiveness in differentiating subtle structural variations in the JAK and FGFR families.

Receptor.AI’s 5-Level Compound Selectivity Platform supports the systematic development of selective compounds for difficult targets, integrating AI-driven prediction with layered selectivity refinement. This approach contributes to ongoing efforts in precision medicine and the rational design of more targeted therapies.

‍

Link to the full article