Fig. 4: Determining and validating core genes decisive for classification. | Nature Machine Intelligence

Fig. 4: Determining and validating core genes decisive for classification.

From: Predicting the prevalence of complex genetic diseases from individual genotype profiles using capsule networks

Fig. 4

a, The distribution of coupling coefficients between primary capsule 5 and phenotype capsule ALS for all genes. The red dashed line indicates the 95th percentile. A total of 922 genes whose coupling coefficients are above the 95th percentile are selected as core genes decisive for classification. The vertical coordinates adopt scientific notation (×106). b, Test accuracy distribution of using 922 randomly chosen genes as input for DiseaseCapsule model (repeat 1,000 times), while the other genes are masked (set as zero). The red dashed line indicates the test accuracy of using 922 core genes as input. c, Heat map for average coupling coefficient matrices (test data) using 922 core genes as input.

Source data

Back to article page