Sophont predicts gene activity from slides. What is established?

On 20 October, startup Sophont announced Holon-S1. It predicts where genes and biological programs may be active in a tumour from a routine H&E pathology slide. Sophont offers an interactive example and invites research collaborations. Its company-authored technical report, dated 13 October, describes retrospective evaluations and biomarker case studies. A prediction from an image is not a direct measurement of gene expression. These materials do not establish that using the model improves treatment choices or survival.

The tools address different steps. Holon-S1 studies tissue images to generate research hypotheses; OpenOnco's free, MIT-licensed browser tool displays source-linked oncology rules and alternative tumour-board plan drafts, but cannot read slides or measure gene expression. OpenOnco's code and rules are public, although clinical content remains under review. On 21 October, Sophont's product page did not publish a Holon-S1 source-code licence or price; access is offered through contact. Neither tool replaces pathology, molecular testing or clinician judgement, and there is no head-to-head clinical outcomes comparison. This article does not replace a qualified physician.