Multi-target protein-chemical interaction prediction using task-regularized and boosted multi-task learning

摘要: Interactions between proteins and small-molecule chemicals modulate many protein functions biological processes, identifying these interactions is a crucial step in modern drug discovery. Supervised learning methods for predicting protein-chemical (PCI) have been widely studied, but their performance largely limited by insufficient availability of binding data proteins. In addition, complex diseases such as Alzheimer's disease cancers are found associated with multiple target Chemicals that selectively only one unable to effectively conquer diseases. this paper we propose two multi-task (MTL) algorithms active compounds related the same diseases, some which may very few examples. first method optimize likelihood compound features Gaussian prior, while second boosts using number independent boosting classifiers. Experimental studies demonstrate significant improvement our MTL over baseline methods. Our also able accurately identify promiscuous interact