MVKTrans: Multi-View Knowledge Transfer for Robust Multiomics Classification

TL;DR

The multi-view knowledge transfer learning (MVKTrans) framework is proposed, which transfers intra- and inter-omics knowledge in an adaptive manner by reviewing data heterogeneity and suppressing bias transfer, thereby enhancing classification performance.

Abstract

The distinct characteristics of multiomics data, including complex interactions within and across biological layers and disease heterogeneity (e.g., heterogeneity in etiology and clinical symptoms), drive us to develop novel designs to address unique challenges in multiomics prediction. In this paper, we propose the multi-view knowledge transfer learning (MVKTrans) framework, which transfers intra- and inter-omics knowledge in an adaptive manner by reviewing data heterogeneity and suppressing bias transfer, thereby enhancing classification performance. Specifically, we design a graph contrastive module that is trained on unlabeled data to effectively learn and transfer the underlying intra-omics patterns to the supervised task. This unsupervised pretraining promotes learning general and unbiased representations for each modality, regardless of the downstream tasks. In light of the varying discriminative capacities of modalities across different diseases and/or samples, we introduce an adaptive and bi-directional cross-omics distillation module. This module automatically identifies richer modalities and facilitates dynamic knowledge transfer from more informative to less informative omics, thereby enabling a more robust and generalized integration. Extensive experiments on four real biomedical datasets demonstrate the superior performance and robustness of MVKTrans compared to the state-of-the-art. Code and data are available at https://github.com/Yaolab-fantastic/MVKTrans.

Figures (3)

• Figure 1: Framework of MVKTrans. (a) The proposed method mainly comprises the following modules. i) Sample-similarity graphs construction. ii) Intra-omics KT: Using pre-trained parameters as initialization, GATs are employed to generate representations and produce omics-specific label distributions, followed by self-attention blocks to prioritize within-omics information. Auxiliary classifiers (AC) are trained to assist in learning more representative features. iii) Inter-omics KT: Cross-omics attention is incorporated with cross-omics distillation (CD) to capture interactions and reconcile distribution disparities among omics. iv) Optimized features are concatenated to make a final prediction. (b-d) illustrates the details of graph contrastive pretraining, self-attention and cross-omics attention, and cross-omics distillation, respectively.
• Figure 2: Performance comparison of different omics combinations.
• Figure 3: Robustness evaluation under different missing rates. Mean and $95\%$ confidence interval (shaded area) are shown.