Peer-reviewed work, software and congress contributions from the last few months. The full record lives on the Publications page.
Over-representation and gene set enrichment analysis (ORA and GSEA) on the chemical-gene interactions curated by the Comparative Toxicogenomics Database: given a gene list from an exposure study, which chemicals does it point to. Available now in the Bioconductor development branch, BiocManager::install("ctdR", version = "devel"), and in the release branch from Bioconductor 3.24.
A retrospective multicentre evaluation of five open-source language models translating free-text Italian mammography reports into structured BI-RADS categories, including whether an ensemble improves on any single model.
A multi-agent system that applies the ACMG/AMP criteria to germline BRCA1/2 variants, compared with a molecular geneticist on the 82 variants classified without ClinGen gene-specific rules, so that both work under the same rules. Substantial agreement (Cohen's kappa 0.70 to 0.84 depending on how the criteria are combined); even the best configuration misses 2 of 43 pathogenic variants. A pilot study, one system, one geneticist, one centre: positioned as decision support rather than replacement.
Transcript-level evidence resolving a novel BRCA1 exon 15-18 duplication, supporting its reclassification as likely pathogenic where DNA analysis alone was inconclusive.
A method comparison between chemiluminescent immunoassay and automated LC-MS/MS for 25-hydroxyvitamin D, quantifying how the choice of platform moves patients across the clinical thresholds for deficiency. The underlying study was published in Diagnostics on 9 August 2026.
Two high-throughput platforms measuring 25-hydroxyvitamin D on the same 1064 paired outpatient samples disagree enough to matter: the immunoassay flags deficiency in 546 samples against 312 for mass spectrometry, so the choice of platform, not the patient, decides a large share of the supplementation decisions.