Supplementary file of ‘Matrix factorization based data fusion for predicting lncRNA-disease associations’

1 Datasets description Ten heterogeneous relational data sources having direct or indirect relevance toward lncRNA-disease association are collected for experiments. The sources of these data sources and statistics of these data sources are listed in Table S1 and Table S2. 2 Optimizing G, S and W This section elaborates on how to iteratively optimize G, S and W in the objective function of MFLDA. Before elaborating on the updating rule, we introduce the Lagrangian multipliers {λi}i=1 for Gi ≥ 0, and reformulate the objective function of MFLDA as follows: min G≥0 L(G,S,W,λ) = ∑ Rij∈R Wijtr(R T ijRij −2Gj RijGiSij + Gi GiSijGj GjSij)