Multi-Modal Deep Learning-Based Model to Predict Burkitt Lymphoma Recurrence
Avery C Maytin1,2, Jessica A Patricoski-Chavez1,2, Ari Pelcovits3
1Center for Computational Molecular Biology, Brown University, Providence, RI 02912.
Summary
Researchers developed a deep learning model, BLIMP, to predict Burkitt Lymphoma (BL) recurrence using multi-modal data. This model shows promise for improving outcomes in aggressive B-cell non-Hodgkin lymphoma patients.
Area of Science:
- Oncology
- Bioinformatics
- Computational Biology
Background:
- Burkitt Lymphoma (BL) is an aggressive B-cell non-Hodgkin lymphoma with a poor prognosis for recurrent cases.
- Currently, there is a lack of predictive models for BL recurrence, despite well-characterized disease pathology.
Purpose of the Study:
- To develop and evaluate a deep learning model for predicting the recurrence of Burkitt Lymphoma.
- To assess the model's performance against traditional machine learning approaches.
Main Methods:
- Developed BLIMP (Burkitt Lymphoma multI-Modal recurrence Predictor), a deep learning model.
- Integrated clinical, gene expression, and mutation data from 184 patients.
- Validated the model on a held-out testing set and performed explainability analysis.
Main Results:
- The BLIMP model achieved an Area Under the Curve (AUC) of 0.788 on the testing set.
- Deep learning approach outperformed traditional machine learning models in predicting BL recurrence.
- Explainability analysis confirmed that predictive features align with known BL pathophysiology.
Conclusions:
- Deep learning models utilizing genomic data are effective for predicting BL recurrence.
- BLIMP demonstrates potential for improving recurrence prediction and guiding future research in BL.
- This approach highlights the utility of multi-modal data integration in cancer recurrence modeling.