Figure 2.

Sliding Window Association Test (SWAT) for genetic variants. (A) Inside view of a sliding window that traverses the entire genome sequence to find a location that is associated with a specific phenotype. A CNN consisting of a convolutional layer with a kernel size of 10, a pooling layer with a maximum pool size of 2, a fully connected layer of 64 nodes, and an output layer with softmax activation was used. (B) Framework to calculate phenotype influence scores of SNPs. We divided the whole genome into 134 955 fragments, each with 40 SNPs. To calculate a phenotype influence score for each of the 40 SNPs included in one fragment, we used an overlapping window approach and CNN. w is the number of SNPs in the fragment and Sk is the kth SNP in the fragment.