Fig. 1
From: PrognosiT: Pathway/gene set-based tumour volume prediction using multiple kernel learning
Our proposed MKL algorithm, named PrognosiT, which takes gene expression profiles of patients, denoted as \(\mathbf {X}\), tumour volumes of patients, denoted as \(\varvec{y}\), and a pathway/gene set collection as its input. Then, it calculates distinct kernel matrices, denoted as \(\mathbf {K}_{1}, \dots , \mathbf {K}_{P}\), for input pathways/gene sets on gene expression partitions, denoted as \(\mathbf {X}_{1}, \dots , \mathbf {X}_{P}\), formed from the input matrix of gene expression profiles. Followingly, to have a kernel matrix between pairs of patients which is denoted as \(\mathbf {K}_{\eta }\) and which carries more information, multiple kernel matrices are combined with a weighted sum. The resulting kernel matrix is later used to learn a function, denoted as f, to predict tumour volumes of out-of-sample cancer patients