Consensus queries in ligand-based virtual screening experiments

Berenger, Francois; Vu, Oanh; Meiler, Jens

doi:10.1186/s13321-017-0248-5

Journal of Cheminformatics

Table 3 Median AUC and PM values with their median absolute deviations

From: Consensus queries in ligand-based virtual screening experiments

Metric	Csize	Policy	MLQSAR dataset			NRLIST dataset
Metric	Csize	Policy	MACCS	ECFP4	UMOP2D	MACCS	ECFP4	UMOP2D
AUC	2	Oppo	0.542±0.059^✠▼	0.548±0.052^✠	0.571±0.055	0.643±0.119	0.699±0.124	0.696±0.131
		Opti	0.514±0.061^▼	0.542±0.046^✠◊	0.564±0.057^◊	0.624±0.115	0.698±0.129^◊	0.700± 0.142^◊
		Real	0.549±0.061^◊	0.052±0.053^✠◊	0.518±0.060^▼	0.632±0.106	0.696±0.119^◊	0.646±0.105
		Sing	0.544±0.064^◊	0.530±0.053^◊	0.540±0.051	0.581±0.100	0.603±0.104	0.598±0.122
	5	Oppo	0.543±0.055	0.530±0.053	0.577±0.057	0.577±0.057	0.814±0.088	0.799±0.992
		Opti	0.460±0.073^▼	0.536±0.040	0.569±0.051^◊	0.625±0.092	0.802±0.111	0.802±0.098
		Real	0.569±0.058^◊	0.562±0.052^◊	0.496±0.057^▼	0.675±0.086	0.774±0.087	0.709±0.092
		Sing	0.536±0.065	0.521±0.051	0.536±0.053	0.574±0.100	0.600±0.106	0.598±0.123
	10	Oppo	0.569±0.047	0.586±0.047	0.607±0.051	0.847±0.061	0.873±0.055	0.856±0.060
		Opti	0.435±0.090^▼	0.554±0.045	0.591±0.043^◊	0.601±0.078	0.853±0.081	0.841±0.084
		Real	0.571±0.065^◊	0.578±0.044	0.495±0.053^▼	0.705±0.078	0.803±0.083	0.738±0.081
		Sing	0.532±0.067	0.525±0.054	0.537±0.049	0.573±0.099	0.597±0.107	0.595±0.125
\(\hbox {PM}_{10\%}\)	2	Oppo	0.588±0.059^✠	0.615±0.058	0.615±0.074	0.912±0.088	0.933±0.067	0.933±0.067
		Opti	0.542±0.071^▼	0.609±0.053^◊	0.621±0.073^✠◊	0.875±0.116	0.941±0.059^◊	0.930±0.070^◊
		Real	0.588±0.066^✠◊	0.593±0.074^✠◊	0.536±0.074^▼	0.896±0.104^◊	0.933±0.067^◊	0.758±0.140
		Sing	0.575±0.077^◊	0.585±0.075	0.585±0.084	0.826±0.155	0.870±0.130	0.832±0.163
	5	Oppo	0.600±0.058	0.635±0.065	0.651±0.060	0.950±0.050	0.968±0.032	0.959±0.041
		Opti	0.452±0.075^▼	0.613±0.059^◊	0.654±0.055^◊	0.804±0.137^▼	0.979±0.021	0.969±0.031
		Real	0.612±0.073^◊	0.623±0.068^◊	0.523±0.068^▼	0.899±0.101	0.955±0.045	0.763±0.160^▼
		Sing	0.569±0.086	0.568±0.079	0.579±0.084	0.817±0.152	0.866±0.130	0.830±0.161
	10	Oppo	0.637±0.047	0.690±0.045	0.688±0.043	0.964±0.036	0.968±0.032	0.956±0.040
		Opti	0.405±0.099^▼	0.660±0.051	0.674±0.041^◊	0.726±0.170^▼	0.973±0.027	0.969±0.031
		Real	0.617±0.063^◊	0.635±0.074	0.528±0.063^▼	0.896±0.090	0.951±0.049	0.724±0.146^▼
		Sing	0.571±0.082	0.576±0.078	0.576±0.079	0.809±0.158	0.862±0.134	0.830±0.155

‘csize’ stands for consensus size. A ‘✠’ indicates that a distribution of peformance metric values is not significantly different from the one of the single policy (Kolmogorov–Smirnov test with p-value \(\ge 0.05\)). A ‘◊’ indicates that a distribution of peformance metric values is not significantly different from the one of the opportunist policy. A ‘▼’ indicates performance worse than the single policy

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com