Comparison of DNA methylation based classification models for precision diagnostics of central nervous system tumors.

in NPJ precision oncology by Quynh T Tran, Alex Breuer, Tong Lin, Ruth Tatevossian, Sariah J Allen, Michael Clay, Larissa V Furtado, Mark Chen, Dale Hedges, Tylman Michael, Giles Robinson, Paul Northcott, Amar Gajjar, Elizabeth Azzato, Sheila Shurtleff, David W Ellison, Stanley Pounds, Brent A Orr

TLDR

  • The study aimed to create a tool that could help doctors classify brain tumors more accurately. They used a computer program called a neural network to analyze DNA samples from brain tumors. They then compared the performance of this program to two other methods, k-nearest neighbor and random forest. They found that the neural network was the best at classifying the brain tumors and was also the most resistant to changes in the quality of the DNA samples. This study highlights the potential of using computer programs to help doctors make more accurate diagnoses and improve patient outcomes.

Abstract

As part of the advancement in therapeutic decision-making for brain tumor patients at St. Jude Children's Research Hospital (SJCRH), we developed three robust classifiers, a deep learning neural network (NN), k-nearest neighbor (kNN), and random forest (RF), trained on a reference series DNA-methylation profiles to classify central nervous system (CNS) tumor types. The models' performance was rigorously validated against 2054 samples from two independent cohorts. In addition to classic metrics of model performance, we compared the robustness of the three models to reduced tumor purity, a critical consideration in the clinical utility of such classifiers. Our findings revealed that the NN model exhibited the highest accuracy and maintained a balance between precision and recall. The NN model was the most resistant to drops in performance associated with a reduction in tumor purity, showing good performance until the purity fell below 50%. Through rigorous validation, our study emphasizes the potential of DNA-methylation-based deep learning methods to improve precision medicine for brain tumor classification in the clinical setting.

Overview

  • The study aimed to develop three classifiers, a deep learning neural network (NN), k-nearest neighbor (kNN), and random forest (RF), to classify central nervous system (CNS) tumor types based on reference series DNA-methylation profiles. The models' performance was validated against 2054 samples from two independent cohorts. The study's primary objective was to compare the robustness of the three models to reduced tumor purity, a critical consideration in the clinical utility of such classifiers. The study emphasizes the potential of DNA-methylation-based deep learning methods to improve precision medicine for brain tumor classification in the clinical setting.

Comparative Analysis & Findings

  • The NN model exhibited the highest accuracy and maintained a balance between precision and recall. The NN model was the most resistant to drops in performance associated with a reduction in tumor purity, showing good performance until the purity fell below 50%.

Implications and Future Directions

  • The study highlights the potential of DNA-methylation-based deep learning methods to improve precision medicine for brain tumor classification in the clinical setting. Future research should focus on developing more robust classifiers that can handle reduced tumor purity and incorporating additional clinical and molecular features to improve the accuracy of the models.
Read Full Article