Splicing-based Predictive Learning for Individual Chemotherapy Evaluation in Colorectal Cancer

Colorectal Cancer, Colorectal Cancer Recurrent, Colorectal Cancer Stage II, Colorectal Cancer Stage III

Chemotherapy, Adjuvant, Response, Splicing, Prediction

Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide. Although adjuvant chemotherapy improves survival after curative resection, its efficacy varies widely among patients. The absence of reliable predictive biomarkers often leads to overtreatment or undertreatment. This study aims to develop a machine learning-based predictive model for adjuvant chemotherapy response using tumor-derived alternative splicing signatures. By integrating RNA-seq data, splicing isoform and clinical outcomes, this study seeks to identify molecular predictors of treatment response and recurrence risk after surgery.

Colorectal cancer (CRC) remains a major global health burden, with adjuvant chemotherapy representing the standard of care after curative resection. However, patient responses to therapy vary widely, and no validated molecular model currently guides adjuvant treatment selection. Recent studies suggest that aberrant alternative splicing-rather than gene-level expression alone-plays a crucial role in shaping chemotherapy sensitivity and tumor recurrence. Yet, these complex transcriptomic variations are often missed by standard differential expression analyses. The ASPAIRE framework (Alternative Splicing and Predictive mAchIne learnIng for Response Evaluation) applies advanced computational modeling to capture multidimensional splicing features from RNA-seq data and transform them into clinically actionable predictions. In this research effort, the investigators will leverage machine learning to predict adjuvant chemotherapy response for CRC. The research plan will employ three phases: 1. Identification of alternative splicing patterns associated with adjuvant chemotherapy response through RNA sequencing and computational feature extraction. 2. The investigators will then develop an assay based on reverse transcription-quantitative polymerase chain reaction (RT-qPCR) and train a machine-learning model to predict chemotherapy response. 3. The investigators will independently validate the assay. This assay is provisionally termed " SPLICE " (Splicing-based Predictive Learning for Individual Chemotherapy Evaluation in Colorectal Cancer) and will be tested for disease free survival up to five years after treatment. At the end of this study, this assay will have been developed and validated to help clinical decision-making by predicting both disease free survival.

No related papers found yet.

United States