Fast and robust imputation for miRNA expression data using constrained least squares

BMC Bioinformatics

Table 1 Handwritten image data results

Metric	FLI	VIPER	scImpute	Regression	Mean	Zeros
(a) Classification results
AUC	\(\sim\)1	\(\sim\)1	.98	.96	.97	.96
\(F_1\)	.99	.99	.98	.93	.87	.83
ACC	.99	.99	.98	.93	.90	.76
(b) Imputation errors
\(\epsilon _{\mu }\)	.15	.15	.35	.71	.59	.68
\(\epsilon _{\sigma }\)	.08	.10	.55	.32	.07	.04
\(\epsilon _{M}\)	.53	.70	6.2	2.4	.93	.81

Time	FLI	VIPER	scImpute	Regression	Mean	Zeros
(c) Imputation time
\(t_\mu\)	.106	11.0	5.21	.126	\(\sim\)0	\(\sim\)0
\(t_\sigma\)	.066	5.47	2.42	.122	\(\sim\)0	\(\sim\)0
\(t_{\text {max}}\)	.346	33.8	14.6	1.23	\(\sim\)0	\(\sim\)0

a Mean values over curves shown in Fig. 2a–c. b Mean values over curves shown in Fig. 2d–f. c Mean (\(t_\mu\)), standard deviation (\(t_\sigma\)), and maximum (\(t_{\text {max}}\)) imputation times (in seconds) over all test patients. In table (a), \(\sim\)1 indicates that the AUC is strictly greater than than .995. In table (c), \(\sim 0\) indicates the imputation time is strictly less than .0005 s

ISSN: 1471-2105