DPI_CDF: druggable protein identifier using cascade deep forest

BMC Bioinformatics

Table 1 Summary of the existing works on druggable protein prediction

Method/tool	Dataset used	Feature set^a	Proposed model^b	Evaluation method^c
DrugMiner [12]	Jamali et al.	AAC, PCP, DPC	NN	5CV
Sun et al. [13]	Jamali et al.	CTD	NN	5CV/IND
GA_Bagging_SVM [5]	Jamali et al.	PAAC, DPC, RC	SVM	5CV
DrugHybrid_BS [14]	Jamali et al.	monoDIKgap, CC, GAAC	SVM	5CV/IND
Yu et al. [15]	Jamali et al. and Yu et al.	DPC, TPC, Dictionary, CTD	CNN, RNN	5CV/IND
XGB DrugPred [16]	Jamali et al.	RAAAC, S-PseAAC, GDPC	XGB	10CV
SPIDER [18]	Jamali et al. and Yu et al.	AAC, CTD, RC, APAAC, PAAC, DPC	SVM	10CV/IND
DPI_CDF (our method)	Jamali et al. and Yu et al.	CPSR, NQLC, HOG-PSSM	CDF	10CV/IND

^aAAC: amino-acid-composition, DPS: dipeptide propensity score, DPC: dipeptide composition, TPC: tripeptide composition, CTD: composition transition and distribution, CC: cross covariance, CPSR: component protein sequence representation, GAAC: grouped AAC, GDPC: grouped DPC, HOG-PSSM: histogram of oriented gradient position specific scoring matrix, NQLC: Normalized qualitative characteristics, APAAC: amphiphilic pseudo AAC, PAAC: pseudo AAC, PCP: physicochemical properties, RAAAC: reduced alphabet amino acid composition, S-PseAAC: serial pseudo amino acid composition
^bNN: neural networks, SVM: support vector machine, CNN-RNN: convolutional-neural-network and recurrent-neural-network, XGB: extreme gradient boosting, CDF: cascade deep forest
^c5-fold cross-validation (5CV); 10-fold cross-validation (10CV); independent test (IND)

ISSN: 1471-2105