Assessment of emerging pretraining strategies in interpretable multimodal deep learning for cancer prognostication.

Background: Deep learning models can infer cancer patient prognosis from molecular and anatomic pathology information. Recent studies that leveraged information from complementary multimodal data improved prognostication, further illustrating the potential utility of such methods. However, current approaches: 1) do not comprehensively leverage biological and histomorphological relationships and 2) make use of emerging strategies to "pretrain" models (i.e., train models on a slightly orthogonal dataset/modeling objective) which may aid prognostication by reducing the amount of information required for achieving optimal performance. In addition, model interpretation is crucial for facilitating the clinical adoption of deep learning methods by fostering practitioner understanding and trust in the technology.

Methods: Here, we develop an interpretable multimodal modeling framework that combines DNA methylation, gene expression, and histopathology (i.e., tissue slides) data, and we compare performance of crossmodal pretraining, contrastive learning, and transfer learning versus the standard procedure.

Results: Our models outperform the existing state-of-the-art method (average 11.54% C-index increase), and baseline clinically driven models (average 11.7% C-index increase). Model interpretations elucidate consideration of biologically meaningful factors in making prognosis predictions.

Discussion: Our results demonstrate that the selection of pretraining strategies is crucial for obtaining highly accurate prognostication models, even more so than devising an innovative model architecture, and further emphasize the all-important role of the tumor microenvironment on disease progression.