Prediction of Intrinsically Disordered Proteins Using Machine Learning Based on Low Complexity Methods

Prediction of intrinsic disordered proteins is a hot area in the field bio-information. Due to high cost evaluating regions protein sequences using experimental methods, we used low-complexity prediction scheme. Sequence complexity this scheme calculate five features for each residue sequence, including Shannon entropy, Topo-logical Permutation entropy and weighted average values two propensities. Particularly, first time that permutation has been applied sequencing. In addition, data preprocessing stage, an appropriately sized sliding window comprehensive oversampling can be improve performance our scheme, ensemble learning algorithms are also verify results before after. The show adding improves algorithm, which MCC value improved from original 0.465 0.526 proving its universality. Finally, compare simulation with those some existing schemes demonstrate effectiveness.