The performance of PINS, consensus clustering (CC), similarity network fusion (SNF), and iCluster+ in discovering subtypes from gene expression data