MOLI

MOLI (Multi-Omics Late Integration) is a deep learning-based software tool for improved drug response prediction by integrating multiple types of omics data. It takes somatic mutation, copy number aberration, and gene expression data as input and learns features for each data type using type-specific encoding sub-networks.

These learned features are then concatenated into a single representation, which is optimized using a combined cost function consisting of a triplet loss and a binary cross-entropy loss. The triplet loss aims to make the representations of responder samples more similar to each other and different from non-responders. In contrast, the binary cross-entropy loss ensures that the representation is predictive of the response values.