FMLRC: Hybrid long read error correction using an FM-index

BMC Bioinformatics

Table 2 After aligning the corrected reads to a reference genome, sensitivity, specificity, and gain were computed

Method	Reads aligned	TP	FN	Sens.	Spec.	Gain	CPU (s)	Mem (GB)
E. coli K12
Canu	58982	657793	1571579	0.2951	0.9987	0.1482	17421	2.99
CoLoRMap	81485	5538038	35474190	0.1350	0.9998	0.1332	137777	23.06
FMLRC	81851	13562639	15069242	0.4737	0.9996	0.4689	11015	4.60
Jabba	75620	15553372	111609	0.9929	0.9999	0.9920	22922	63.87
LoRDEC	81138	3278911	36691424	0.0820	0.9998	0.0808	61305	2.08
LoRMA	81051	1135657	144161	0.8874	0.9999	0.8669	54240	45.72
Sprai	75532	463636	293039	0.6127	0.9999	0.5783	44302	33.53
S. cerevisiae W303
Canu	142765	1562542	5441239	0.2231	0.9992	0.1401	108175	3.33
CoLoRMap	210423	18901871	79065115	0.1929	0.9992	0.1857	2815200	45.49
FMLRC	211270	31849332	49204291	0.3929	0.9991	0.3829	68519	17.77
Jabba	223385	29893606	42670	0.9986	0.9999	0.9967	187968	367.92
LoRDEC	210151	8468872	96577493	0.0806	0.9997	0.0776	212495	3.56
LoRMA	204323	3063164	221583	0.9325	0.9999	0.9176	223358	49.52
Sprai	192670	2013269	3063751	0.3965	0.9996	0.3288	215261	49.52
A. thaliana Ler-0
Canu	574065	12002535	72017120	0.1429	0.9986	0.1030	1301971	10.92
CoLoRMap	1075381	170235345	2056204621	0.0765	0.9983	0.0737	6802359	106.08
FMLRC	1447042	442064624	1164804073	0.2751	0.993	0.2601	708910	16.26
Jabba	813495	320742341	2945173	0.9909	0.9968	0.9724	3641309	333.44
LoRDEC	1113617	78276025	2022520259	0.0373	0.9979	0.0337	1111800	3.42
LoRMA	903298	2217661	2778223	0.4439	0.9986	0.3715	17281259	70.28
Sprai	751684	18960255	30734331	0.3815	0.9996	0.3631	5996657	8.11

For A. thaliana and S. cerevisiae, FMLRC produced more total true positive (corrected loci) than any other method while maintaining competitive sensitivity and gain. Methods with higher average specificity, notably Jabba, often discard a higher proportion of reads, reporting only those with the highest-confidence corrected sequence. FMLRC also requires significantly less CPU time than other hybrid error correction methods, and comparable memory. LoRDEC and FMLRC CPU time and memory results include construction of the BWT

ISSN: 1471-2105