Machine Learning Model Predicts Response to Radiation in Rectal Cancer
Author Information
Author(s): Domingo Enric, Rathee Sanjay, Blake Andrew, Samuel Leslie, Murray Graeme, Sebag-Montefiore David, Gollins Simon, West Nicholas, Begum Rubina, Richman Susan, Quirke Phil, Redmond Keara, Chatzipli Aikaterini, Barberis Alessandro, Hassanieh Sylvana, Mahmood Umair, Youdell Michael, McDermott Ultan, Koelzer Viktor, Leedham Simon, Tomlinson Ian, Dunne Philip
Primary Institution: University of Oxford
Hypothesis
What biological features predict the response of rectal cancer to radiotherapy?
Conclusion
Rectal cancers that respond completely to radiotherapy have immune activation characteristics and low TGFβ signaling, which can be identified using a 33-gene expression signature.
Supporting Evidence
- Three gene expression sets were significantly associated with complete response to radiotherapy.
- The machine learning model showed 90% sensitivity and 86% specificity in predicting response.
- Biological and machine learning signatures indicated similar mechanisms underlying radiation response.
Takeaway
Doctors can use a special test to see if rectal cancer patients will respond well to radiation treatment, helping them choose the best care.
Methodology
The study analyzed pre-treatment biopsies from two cohorts of rectal cancer patients using machine learning and biological profiling to identify predictors of complete response to radiotherapy.
Potential Biases
Potential biases from sample selection and the need for batch correction in data analysis.
Limitations
The study is limited to patients treated with long course chemoradiotherapy and may not apply to other treatment regimens.
Participant Demographics
Patients with rectal cancer treated with neoadjuvant radiotherapy.
Statistical Information
P-Value
p=0.0006
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website