XSAR: Crystallographic SAR From Crude Reactions

Why It Matters

xSAR provides a data‑driven way to rescue false negatives and prioritize fragments with superior affinity, reshaping early‑stage hit discovery in medicinal chemistry.

Key Takeaways

• xSAR quantifies binding motifs via Morgan fingerprint bits
• PBS outperforms Tanimoto similarity in false‑negative recovery
• Virtual screen of 1.7 B compounds yielded nine new crystal binders
• Two GCI hits showed sub‑micromolar affinity, missed crystallographically
• Method highlights limits of crystal‑first fragment optimization

Pulse Analysis

Crystallographic screening of crude reaction mixtures has emerged as a rapid way to identify fragment binders, but its binary nature often discards valuable chemical space. The recent xSAR study builds on earlier work targeting the oncology protein PHIP(2) by translating each fragment into a high‑dimensional Morgan fingerprint, flagging conserved binding bits (CBB) and conserved non‑binding bits (CNB). By aggregating these features into Positive and Negative Binding Scores, the authors created a quantitative SAR framework that can be applied to both binders and non‑binders, addressing the chronic problem of false‑negative hits inherent in crude‑reaction screens.

In a retrospective analysis, applying PBS to the original 957‑compound set uncovered 97 high‑scoring pure compounds, of which 26 were confirmed as binders, effectively doubling the initial hit rate. PBS proved more sensitive than traditional Tanimoto similarity metrics, highlighting its utility for prioritizing fragments that may be overlooked by conventional similarity searches. This quantitative approach also enabled a massive prospective virtual screen of over 1.7 billion Enamine REAL molecules, where PBS/NBS filtering followed by docking produced a focused library of 93 compounds. Although the crystal hit rate was modest (9 binders), the strategy demonstrated scalability and the ability to explore chemical space far beyond what is feasible with experimental screens alone.

The most striking outcome came from orthogonal GCI biophysical assays, which identified 13 binders with measurable affinity, including two sub‑micromolar fragments that were invisible to crystallography. These results underscore a key insight: crystal‑first fragment discovery excels at locating binding poses but may miss the highest‑affinity chemotypes. Integrating quantitative xSAR metrics with complementary biophysical readouts offers a more balanced pipeline, enabling medicinal chemists to capture both structural information and potency early in drug development. Confidence score reflects high‑quality, industry‑relevant analysis.