TGF-β signaling alterations in bevacizumab-treated early-onset colorectal cancer: An artificial intelligence–enabled precision oncology study in diverse populations.

Abstract

3649 Background: The transforming growth factor–β (TGF-β) signaling pathway plays a complex, context-dependent role in colorectal cancer (CRC), functioning in tumor suppression, immune modulation, and metastatic progression. Its interaction with anti-angiogenic therapy remains poorly defined, particularly across age and ancestry groups. Early-onset CRC (EOCRC) may exhibit distinct TGF-β pathway biology. We investigated ancestry-, age-, and treatment-associated patterns of TGF-β signaling alterations and their prognostic relevance in bevacizumab-treated CRC using an artificial intelligence–enabled precision oncology framework. Methods: We performed a retrospective analysis of 2,717 CRC patients from publicly available datasets with integrated genomic, clinical, and treatment annotations. Eligible cases included colorectal adenocarcinoma with sequencing data and documented bevacizumab exposure. Patients were stratified by age ( < 50 years [EOCRC] vs. ≥50 years [LOCRC]), ancestry (H/L vs. non-Hispanic White [NHW]), and treatment status. TGF-β pathway alterations, including mutations in core signaling components (e.g., TGFBR2), were analyzed as categorical variables. Differences in mutation frequencies were assessed using Fisher’s exact test. Overall survival (OS) was evaluated using Kaplan–Meier methods and Cox proportional hazards models. Conversational AI agents facilitated reproducible cohort construction and multi-dimensional querying. Analyses were exploratory. Results: Marked ancestry-, age-, and treatment-specific differences in TGF-β alterations were observed. Among bevacizumab-unexposed EOCRC, H/L tumors demonstrated substantially higher TGF-β alteration frequencies compared with NHW tumors (31.9% vs. 0.2%, p = 2 × 10⁻¹⁶). In NHW EOCRC, bevacizumab exposure was associated with increased TGF-β alterations relative to unexposed tumors (25% vs. 0.2%, p = 2 × 10⁻¹⁶). Similarly, bevacizumab-treated NHW LOCRC exhibited higher TGF-β alteration frequencies than unexposed counterparts (58% vs. 15.4%, p = 2.2 × 10⁻¹⁶). Among bevacizumab-unexposed LOCRC, H/L tumors showed higher TGF-β alterations than NHW tumors (34.2% vs. 18.0%, p = 2.41 × 10⁻⁵). In NHW LOCRC, bevacizumab-exposed tumors had a lower frequency of TGFBR2 alterations compared with unexposed tumors (1.3% vs. 7.1%, p = 0.03). Survival analyses suggested that TGF-β alterations were associated with improved OS in bevacizumab-unexposed NHW LOCRC, with borderline statistical significance (p = 0.08). Conclusions: The enrichment of TGF-β alterations in specific bevacizumab-exposed and ancestry-defined subgroups highlights potential interactions between angiogenic therapy and TGF-β pathway biology. These findings highlight the importance of incorporating ancestry-aware frameworks to advance precision oncology.