Bioptimus SDK 1.4.0 adds gene signature scoring on M-Optimus spatial gene expression predictions, with one sided and two sided signatures, batched linear scoring and per tile score persistence. It also adds axis aware reducers shared by slide pooling and gene reduction, a total count normalizer with cpm and cp10k recipes, and a base_mpp fallback for slides missing resolution metadata. The release removes the string pool= argument and changes TPMNormaliser to require gene lengths, both flagged as breaking.
Our readPharma and research teams can now score biomarker gene signatures across tiles and cohorts inside the vendor SDK instead of building that step themselves. The breaking changes mean existing pipelines on 1.3.x need code updates before upgrading.