RNA evidence supports clinical stage antibody drug targets in oncology

Background: The selection of efficacious and safe antibody-based oncology targets depends on tumor specificity and normal-tissue expression, and remains challenging. Bulk RNA-seq can identify tumor-enriched surface genes, but may obscure tumor specificity and normal-cell expression patterns that are relevant to the therapeutic window. We evaluated whether single-cell RNA-seq-derived expression features improve retrospective prioritization of clinically pursued antibody targets. Methods: We developed a bulk and single-cell RNA expression-based ranking framework across six solid tumor indications. 4,532 candidate surfaceome targets were scored using tumor Specificity, normal Safety, and blood Sink features derived from bulk RNA-seq, single-cell RNA-seq, or a combined modality. Ranking performance was evaluated against a curated validation set of tumor-associated antibody-based oncology targets using logistic regression, with odds ratios (ORs) reported per one-standard-deviation increase in score. Results: Higher Specificity scores were associated with clinical-stage target selection across all modalities. In cross-indication meta-analysis, single-cell Specificity showed higher ORs than bulk Specificity across clinical phase thresholds, including Phase [≥] 1 (OR 2.56 versus 1.81), Phase [≥] 2 (OR 2.77 versus 1.88), Phase [≥] 3 (OR 3.53 versus 2.04) and Approved targets (OR 3.51 versus 1.74). Combined bulk-and-single-cell Specificity showed intermediate enrichment. Specificity score alone did not consistently show a significant single-cell advantage over bulk in paired indication-level comparisons, except for Phase [≥] 3 targets. After adding normal-cell Safety and blood Sink features, the single-cell general score showed the strongest enrichment across thresholds (Phase [≥] 1 OR 3.61, Phase [≥] 2 OR 4.16, Phase [≥] 3 OR 5.57, and Approved targets OR 4.25). This general score also showed consistently significant single-cell advantage over bulk in paired indication-level comparisons. Safety contributed most strongly in the single-cell modality, increasing meta-analytic ORs by an average of ~36% across thresholds compared to Specificity alone. Performance varied by indication, with breast cancer showing weaker Specificity-based enrichment. Conclusion: Expression-derived ranking features are retrospectively enriched among clinically pursued tumor-associated antibody targets. Single-cell evidence is most informative when tumor Specificity is combined with normal-cell Safety information, supporting its use as an interpretable component of early antibody target prioritization. These findings are in alignment with recent evidence that single-cell RNA-based cell-type specificity supports clinical-stage targets and demonstrate that expression-driven ranking can systematically recover clinically relevant oncology targets.