Table 1.

Sequence-based protein domain identification methods.

Category	Method	Description	Year	URL	Reference
Homology-based	CHOP	Search target sequences against PDB, Pfam-A, and SWISS-PROT to find templates.	2004	http://www.rostlab.org/services/CHOP/	[6]
	DomPred	Combine homology and secondary structure element alignment to find templates.	2005	http://bioinf.cs.ucl.ac.uk/software.html	[15]
	SSEP-Domain	Based on the secondary structure elements alignment and profile-profile alignment.	2006	http://www.bio.ifi.lmu.de/SSEP/	[16]
	ThreaDom	Deduce domain boundary locations based on multiple threading alignments.	2013	http://zhanglab.ccmb.med.umich.edu/ThreaDom/	[18]
	CLADE	Identifie domains by a multi-source strategy which combining multiple HMMs profile.	2016	http://www.lcqb.upmc.fr/CLADE	[13]
	MetaCLADE	A multi-source domain annotation tool for metagenomic dataset.	2018	http://www.lcqb.upmc.fr/metaclade	[14]
Ab initio methods	Domain Guess by Size	Detect domain boundaries based on the distributions of chain and domain lengths.	2000		[26]
	CHOPnet	Feed-forward neural network that uses amino acid composition and secondary structure and solvent accessibility as features.	2004		[27]
	PPRODO	Feed-forward neural network that uses position-specific scoring matrix (PSSM) generated by PSI-BLAST as features.	2005	http://gene.kias.re.kr/~jlee/pprodo/	[28]
	DOMpro	RNN uses secondary structure and solvent accessibility as features.	2005	http://www.igb.uci.edu/servers/psss.html	[31]
	KemaDom	Combine three SVM classifiers that use different features as inputs to predict domain boundaries.	2006	http://www.iipl.fudan.edu.cn/lschen/kemadom.htm	[30]
	DomainDiscovery	SVM uses inter-domain linker index, PSSM, secondary structural, and solvent accessibility as features.	2006		[33]
	IGRN	An improved general regression network model that is trained by the information of PSSM, interdomain linker index, secondary structure, and solvent accessibility.	2008		[32]
	DomSVR	Sequence is encoded by physicochemical and biological properties. SVR uses encoded sequence to predict domain boundary.	2010		[35]
	DoBo	SVM uses evolutionary domain boundary signals embedded in homologous proteins as input features.	2011	http://sysbio.rnet.missouri.edu/dobo/	[34]
	DROP	An SVM to predict domain linkers using 25 optimal features selected from a set of 3000 features.	2011	http://tuat.ac.jp/~domserv/DROP.html	[37]
	DomHR	Identify domain boundaries in proteins by defining the edge of domain and boundary regions as a hinge region.	2013	http://cal.tongji.edu.cn/domain/	[39]
	PDP-CON	Combine predicted results from six single domain boundary prediction methods.	2016	https://cmaterju.org/cmaterbioinfo/	[38]
	ConDo	Use long-range, coevolutionary features to train neural networks.	2018	https://github.com/gicsaw/ConDo.git	[40]
	DNN-Dom	Combine CNN and BGRU to predict domain boundary by combining amino acid composition information, PSSM, solvent accessibility, and secondary structure. A balanced Random Forest is used to solve the imbalance samples problem.	2019	http://isyslab.info/DNN-Dom/	[42]
	DeepDom	Use sequences information encoded by physical–chemical properties to train a bidirectional LSTM model to predict domain boundaries.	2019	https://github.com/yuexujiang/DeepDom	[41]
	FuPred	Predict protein domain boundaries using predicted contact maps generated by ANN.	2020	https://zhanglab.ccmb.med.umich.edu/FUpred	[44]