Fig. 2. Performance comparison of multi-view data prediction via M²AE and single-view data prediction via Attentive Encoders in Dataset 1 (n = 5 experiments for each model).

Means of evaluation metrics from different experiments are shown in the figure. MGS + Host characteristics + Dietary refers to M²AE with three types of views combining MGS, host characteristics, and dietary features data. MGS + Host characteristics refers to M²AE with two types of views combining gut microbiomes and host characteristics. MGS + Dietary refers to M²AE with two types of views combining gut microbiomes and dietary features. Host characteristics + Dietary refers to M²AE with two types of views combining host characteristics and dietary features. MGS, host characteristics and dietary features refer to the view-specific attentive encoders trained with single-view gut microbiomes, host characteristics, and dietary features.