DMoVGPE: predicting gut microbial associated metabolites profiles with deep mixture of variational Gaussian Process experts

Table 3 Comparative Analysis

Dataset	Disease-Label-Based Split (DMoVGPE)	Random Data Splitting	Increased Expert Count (Double Experts)
MFGM	0.566 ± 0.042	0.556 ± 0.038	0.554 ± 0.038
ASD	0.703 ± 0.104	0.695 ± 0.089	0.587 ± 0.085
INFANTS_DIABETES	0.728 ± 0.044	0.668 ± 0.053	0.710 ± 0.062
PRETERMS	0.797 ± 0.033	0.778 ± 0.041	0.773 ± 0.036
SINHA_CRC	0.612 ± 0.059	0.593 ± 0.033	0.552 ± 0.024
ADENOMAS	0.667 ± 0.058	0.628 ± 0.054	0.604 ± 0.069
BIO_ML	0.749 ± 0.024	0.749 ± 0.024	0.744 ± 0.026
IBD_RELATIVES	0.790 ± 0.016	0.763 ± 0.052	0.767 ± 0.039
IBS	0.465 ± 0.051	0.462 ± 0.061	0.494 ± 0.051
CRC	0.555 ± 0.034	0.561 ± 0.035	0.521 ± 0.031
GC	0.669 ± 0.063	0.528 ± 0.052	0.596 ± 0.037
ESRD	0.634 ± 0.036	0.609 ± 0.037	0.586 ± 0.035
IBD	0.857 ± 0.027	0.821 ± 0.026	0.838 ± 0.024
IBDMDB	0.674 ± 0.020	0.521 ± 0.065	0.459 ± 0.074

The bolded values indicate the best results among the three expert allocation strategies

ISSN: 1471-2105