A fair experimental comparison of neural network architectures for latent representations of multi-omics for drug response prediction

Table 4 Mean AUPRC on external test set

Drug	Omics stacking	MOLI	Super.FELT	Early integration	OmiEmbed	MOMA	PCA
Gemcitabine TCGA	0.581 ± 0.074	0.535 ± 0.085	0.502 ± 0.082	0.513 ± 0.086	0.462 ± 0.095	0.389 ± 0.040	0.421 ± 0.081
Gemcitabine PDX	0.510 ± 0.132	0.424 ± 0.038	0.457 ± 0.055	0.362 ± 0.108	0.466 ± 0.100	0.414 ± 0.068	0.481 ± 0.131
Cisplatin	0.942 ± 0.027	0.950 ± 0.007	0.963 ± 0.009	0.932 ± 0.004	0.908 ± 0.009	0.952 ± 0.006	0.964 ± 0.012
Docetaxel	0.560 ± 0.051	0.590 ± 0.021	0.565 ± 0.024	0.491 ± 0.034	0.544 ± 0.049	0.578 ± 0.060	0.523 ± 0.089
Erlotinib	0.440 ± 0.075	0.410 ± 0.158	0.223 ± 0.024	0.428 ± 0.106	0.294 ± 0.079	0.369 ± 0.146	0.427 ± 0.138
Cetuximab	0.125 ± 0.018	0.141 ± 0.086	0.126 ± 0.039	0.108 ± 0.028	0.101 ± 0.018	0.148 ± 0.065	0.110 ± 0.032
Paclitaxel	0.256 ± 0.147	0.191 ± 0.077	0.147 ± 0.028	0.120 ± 0.016	0.135 ± 0.007	0.172 ± 0.047	0.163 ± 0.094
Mean Rank	2.43	2.86	4.00	5.43	5.57	3.86	3.86

Best results are shown in bold and second best are italic. The values represent the means and standard deviations over five iterations

ISSN: 1471-2105