AI Science Tracker

Contrastive-learning biomarker discovered from phase 2 data retrospectively generalizes to phase 3 immunotherapy outcomes

AstraZeneca; Tempus AI · 12 May 2025 · medicine

Discovery

Contrastive-learning biomarker discovered from phase 2 data retrospectively generalizes to phase 3 immunotherapy outcomes

The Predictive Biomarker Modeling Framework (PBMF), an ensemble neural network trained with a treatment-versus-control contrastive objective, identified a composite clinicogenomic biomarker using phase 2 POPLAR data for atezolizumab versus docetaxel. When applied retrospectively to the phase 3 OAK trial, its biomarker-positive group had a treatment-versus-control overall-survival hazard ratio of 0.59 (95% CI 0.47–0.74); an interpretable distilled decision tree produced HR 0.55 (95% CI 0.43–0.70). The tree used baseline ctDNA maximum allele frequency, lesion diameter, and mutation status in KMT2D, TSC1, ATM, PDGFRA and LRP1B.

Technology: Predictive Biomarker Modeling Framework (PBMF), an ensemble of neural networks using a contrastive treatment/control objective; knowledge-distilled decision tree

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

Result, AI contribution and limits

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

  • Publication status: Peer-reviewed paper
  • Test type: retrospective or benchmark
  • Tested: The Predictive Biomarker Modeling Framework (PBMF), an ensemble neural network trained with a treatment-versus-control contrastive objective, identified a composite clinicogenomic biomarker using phase 2 POPLAR data for atezolizumab versus docetaxel.
  • Still unknown: Retrospective reanalysis; the biomarker has not prospectively guided patient selection or treatment. The outcome comparisons do not establish benefit from applying PBMF clinically. The authors call for prospective validation and note potential cohort comparability limitations. The identified feature set includes previously studied biomarkers; the paper’s novel result is the composite PBMF-derived predictive subgroup and retrospective generalization, not discovery of each individual feature.
  • AI role: PBMF searched clinicogenomic features for a treatment-specific predictive subgroup, using phase 2 POPLAR training data and contrasting atezolizumab against docetaxel outcomes. Researchers transferred the derived model/biomarker to phase 3 OAK data and distilled it into an interpretable tree. This was a retrospective analysis of trial data, not prospective trial stratification.

Evidence sources: https://ellykipkogei.io/assets/pdf/1-s2.0-S1535610825001308-main.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review scope and date are not recorded; source checked does not establish full-text review.

Citations: in the top 1% of works of the same type, year and subfield (OpenAlex work metrics, 65 citations; as of 15 Nov 2026). Attention, not a measure of correctness; not part of the score.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Follow-up and relationships

No follow-up recorded; this is not a claim that no follow-up exists.

Follow-up review scope and date are not recorded.

No source-supported relationships recorded.

Open in the full catalog for filters, charts and related discoveries.