TABLE 2.

Representative diseases-, phenotypes-, genes-specific variants impact predictors.

Characteristic category	Name	Type of variants	Targeted disease/phenotype/gene	# of genes	Website	Distribution (web-server/stand-alone)	First publication	Programming language	Algorithm/model	Features	Dataset for modeling	Classification index	Classification	Additional data	Publication
Meta-predictor	VIPPID (Variant Impact Predictor for PIDs)	Missense	Primary immunodeficiency (PID) diseases	146	https://mylab.shinyapps.io/VIPPID/	Web and stand-alone	April 2022	Perl, R	Conditional Inference Forest	85 features including AA, exonic, protein structural, conservation, and 20 pre-existing prediction tools	4,865 disease-associated variants from Asian Primary Immunodeficiency Diseases (RAPID) database, HGMD and ClinVar; 4,237 neutral variants from gnomAD	Classifier	Pathogenic/non-pathogenic	26 reviewed P/LP variants of known PID pathogenic genes from 1318 patients cohort and 39 validated in-house variants	Fang et al. (2022)
Meta-predictor	CanPredict	Missense	Cancer	—	http://www.canpredict.org/or http://www.cgl.ucsf.edu/Research/genentech/canpredict/, both are not accessible	—	May 2007	R	RF	SIFT, Pfam-based LogR.E-value and GO Similarity Score (GOSS) metrics	—	Classifier	Likely cancer/likely non-cancer/not determined	—	Kaminker et al. (2007)
Meta-predictor	PolyPhen-HCM	Missense	Hypertrophic cardiomyopathy	6	http://genetics.bwh.harvard.edu/hcm/	Pre-computed results	February 2011	—	Naïve bayes classifier	Prediction scores, protein structure comparison score	74 curated variants from literitures and manually classified by Laboratory for Molecular Medicine standard variant-assessment pipeline (41 pathogenic, 26 benign)	Classifier	Pathogenic/benign/no call	—	Jordan et al. (2011)
Meta-predictor	Cadioboost	Missense	Cardiomyopathies and arrhythmias	22	https://www.cardiodb.org/cardioboost/	Pre-computed results	October. 2020	R	2 Adaptive Boosting (Adaboost) classifiers	76 functional features	CM datasets: 356 rare P/LP variants from 9,007 clinical CM patients, 302 rare missense variants in CM genes from 2,090 healthy controls. Inherited arrhythmia dataset: 252 P/LP in arrhythmia-associated genes from ClinVar, 237 rare missense variants in arrhythmia genes from 2,090 healthy controls	Pathogenicity score	Disease-causing/VUS/Benign	4 datasets from ClinVar, HGMD, Oxford Medical Genetics Laboratory (OMGL), a large registry of HCM patients, SHaRe	Zhang et al. (2021)
Multiple features	GENESIS (GENe-specific EnSemble grId Search)	Variants of uncertain clinical significance	Catecholaminergic polymorphic ventricular tachycardia and long QT syndrome (LQTS)	4	https://github.com/rachellea/medgenetics	Stand-alone and pre-computed results	March 2022	Python	Logistic regression and multilayer perceptron model	8 kinds of features including AA features, domain, conservation, rate of evolution, signal-to-noise ratio, and a position-specific scoring matrix (PSSM) score	717 pathogenic variants and 3,164 benign variants curated from literiture	Probabilities of pathogenicity	Pathogenic/VUS/benign	925 VUS classified according to ACMG	Draelos et al. (2022)
Multiple features	CACNA1F-vp	Missense	X-linked incomplete Congenital Stationary Night Blindness (iCSNB)	1	https://github.com/shalawsallah/CACNA1F-variants-analysis	Stand-alone	April 2020	Python	Logistic regression model	Variant-level features and structural features	72 disease-implicated from HGMD or MGDL database, 322 benign variants from gnomAD	Probabilities of pathogenicity	Pathogenic/benign	-	Sallah et al. (2020)
Optimized PON-P2	PON-MMR2	AA substitution	Mismatch repair (MMR)	4	http://structure.bmc.lu.se/PON-MMR2/	Web and stand-alone	September 2015	R	RF	5 features: sequence conservation, physical and biochemical properties of AA	109 pathogenic, 99 neutral, 354 VUS from InSiGHT database and VariBench	Probabilities of pathogenicity	Pathogenic/VUS/benign	354 VUS dataset	Niroula and Vihinen, (2015)
Optimized MAPP	CoDP (Combination of Different Properties of MSH6 protein)	Missense	Lynch syndrome (LS)	1	http://cib.cf.ocha.ac.jp/CoDP/	Web	April 2013	—	Logistic regression model	MSA, phylogenetic tree, structral properties, MAPP, SIFT, PolyPhen2	294 missense variants from InSiGHT, MMRUV, UniProt, dbSNP, ESP, HapMap Project, 1KGP and literature	Probabilities of pathogenicity	Likely LS/Unlikely LS	260 unclassified variants dataset	Terui et al. (2013)
Meta-predictor with MAF as features	DvPred	nsSNVs	Genetic hearing loss (HL)	157	https://github.com/WCH-IRD/DVPred/tree/main/DVPred_score	Stand-alone and pre-computed results	February 2022	Python	Gradient boosting decision tree (GBDT)	65 features include conservation scores, prediction scores, MAF, gene intolerance scores and other features	1,318 P/LP and 4,628 B/LB from China Deafness Genetics Consortium (CDGC), Deafness Variation Database (DVD), ClinVar, HGMD	DvPred score	Deleterious/neutral	463 pathogenic and 454 benign variants from new version of CDGC and ClinVar	Bu et al. (2022)
Meta-predictor	NBDriver	Missense	Cancer	58	https://github.com/RamanLab/NBDriver	Stand-alone	May 2021	Python	RF, extra tress (ET) classifier, generative KDE classifier	3 types of features: one-hot encoding, overlapping k-mers, 27 genomic features	5,265 disease-associated variants from five literatures	Classifier	—	—	Banerjee et al. (2021)
Combination of rule-based and meta-predictor	CancerVar	Exon variants, CNVs, indels	Cancer	1911	https://cancervar.wglab.org/index.php	Web, stand-alone and pre-computed results	May 2022	Python	Semi-supervised generative adversarial network used in scoring method OPAI	12 clinical evidence prediction scores and 23 precomputed scores by other computational tools	13 million variants from 7 cancer knowledgebases	OPAI score	Oncogenic/benign	4 datasets from OncoKB and CIViC, IARC and literatures	Li et al. (2022)

*VUS, variant of uncertain significance.