Back

Phenotype-first covalent fragment screening identifies a synthetic lethal TYMS inhibitor in ATRX-deficient cells

Raguseo, F.; Fellows, E.; de Chiara, C.; Mortishire-Smith, B.; Segura-Bayona, S.; Cawood, E.; Jiang, M.; Subtil, F. T.; Idilli, A.; McCarthy, W.; Howell, M.; Rittinger, K.; MacRae, J. I.; Skehel, M.; Powell, A.; House, D.; Bush, J.; Boulton, S.

2026-06-11 cell biology
10.64898/2026.06.10.731339 bioRxiv
Show abstract

Chemoproteomic mapping of covalent fragment libraries is expanding the ligandable human proteome with direct evidence of cellular target engagement. However, understanding the functional consequences of specific covalent modifications typically requires extensive downstream biological characterisation. Here we present a phenotype-first approach that integrates covalent fragment screening with chemoproteomics and genetic deconvolution in a disease-relevant context. Using isogenic ATRX wild-type and knockout eHAP iCAS9 cells, we screened a library of around 500 cysteine-reactive fragments for differential cell killing and identified a chloroacetamide fragment, PP12, that selectively impairs the viability of ATRX-deficient cells. By combining competitive click-chemoproteomics with genome-wide CRISPR synthetic lethal datasets, we identified thymidylate synthase (TYMS) as a phenotypically relevant target of PP12. Target validation was supported by crystallography, competition with the active-site inhibitor 5-fluorouracil, and impaired dTMP synthesis in cells. Mechanistically, TYMS inhibition induces replication stress that is selectively cytotoxic to ATRX-deficient cells and is dependent on FAM111A and SLFN11. This work establishes a generalisable workflow linking covalent fragment phenotypes to target deconvolution using chemoproteomics and mechanistic validation.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.