Software Daily Summary: What Happened on November 3, 2026

Mira’s clinical evidence varies in coverage by development stage

Mira brings together clinical reports from 23 sources, standardizes drug and disease identifiers, and assigns reports to 13 shared clinical stages.1 The pipeline supplied 289,954 reports for the September 2026 release, compared with 285,213 around the March release.1

The September Clinical Precedence dataset drew on 117,152 reports, including 94,480 clinical-trial reports, and covered 11,418 drug-disease pairs in the platform’s Indications widget.1 Open Targets estimates that roughly one-third of drugs previously listed at Phase III moved to Approval, largely because newer regulatory labels were available.1

At Approval, 55% of indications had at least two supporting reports and 41% had three or more. Phase I evidence mostly relied on one report.1 More than 90% of Approval-stage reports were fully mapped, compared with 55% at Phase I and under 10% at Preclinical.1

Of 47,870 target-disease pairs compared between releases, more than 9,500 early-development pairs gained evidence score, with an average increase of 0.323.1 Around 3,200 pairs in a late-clinical score band lost an average of 0.5. One example was CTLA4-cutaneous melanoma, whose score fell from 0.7 to 0.2 after two ipilimumab trials were mapped to broader melanoma.1

Open Targets says more than 290,000 reports meet Mira’s inclusion criteria, but mapping limitations mean fewer than half contribute to target-disease evidence.1 In June, the team added a large language model framework to extract study purpose, drugs, and indications from ClinicalTrials.gov records. The authors reported more drug-indication pairs and fewer false associations against a manually annotated benchmark.1

Also published

• The Shelf Life of LLM findings Paired Ends
• Cambridge Cheminformatics Network meeting Mac in Chem
• AI: an early warning system for patient safety Pharmafile
• ChemDraft an open-source chemical drawing package Mac in Chem