TABLE 1.

General information on the tools used here

Tool	Method	Training sequence data seta	No. of E. coli sigma factors	Availability	Yr	Reference	No. of citations (Google Scholar)b
BPROM	Weight matrices of different motifs combined with linear discriminant analysis	Positive: Experimentally validated promoters from E. coli (14).	70	Web server http://www.softberry.com/berry.phtml?topic=bprom&group=programs&subgroup=gfindb	2011	33	427
		Negative: Inner regionsof protein-coding ORFs.
bTSSfinder	Position weight matricesfor promoter elements, oligomer frequencies, physicochemical properties as features, and Mahalanobis distance for feature selection and with neural network for classification	Positive: Experimentally validated TSSs from Regulon DB.	24, 28, 32, 38, 70	Stand-alone and Web server http://www.cbrc.kaust.edu.sa/btssfinder/	2016	23	26
		[−200, +51].
		Negative: Genomic regions withno experimental evidence for the presence of TSSs.
BacPP	Weighted rules extracted from neural network	Positive: Regulon DB available promoters.	24, 28, 32 38, 54, 70	Web server http://www.bacpp.bioinfoucs.com/home	2011	17	22
		[−60, +20].
		Negative: randomly generated sequences (with established nucleotide frequencies) and intergenic sequences.
Virtual Footprint	PWMs from different available databases			Web server http://www.prodoric.de/vfp/vfp_promoter.php	2005	36	370
IBBP	Image-based and evolutionary approach which generates “images” (template-image strings that keep features of spatial sequence relationships)	Positive: sigma 70 promoters from Regulon DB.	70 (expand- able approach)	Source code https://github.com/hahatcdg/IBPP	2018	35	1
		[−60, +20].
		Negative: randomly generated from protein-coding sequences.
iPro70-FMWin	22,595 features extracted from sequence and AdaBoost to select the most representatives among then; logistic regression classifier	Positive: Regulon DB annotated promoters.	70	Webserver http://ipro70.pythonanywhere.com/	2019	38	4
		[−60, 20].
		Negative: randomly generated from protein- coding and intergenic region sequences.
70ProPred	Support vector machine using position-specific tendencies of trinucleotide and electron-ion interaction pseudopotentials as features	Positive: promoters from Regulon DB.	70	Webserver http://server.malab.cn/70ProPred/	2017	39	33
		[−60, 20].
		Negative: randomly generated from coding and noncoding sequences.
CNNProm	Convolutional neural networks	Positive: promoters from Regulon DB.	70	http://www.softberry.com/berry.phtml?topic=index&group=programs&subgroup=deeplearn	2017	34	60
		[−60, 20].
		Negative: the opposite chain of randomly selected protein-coding genes.
MULTiPly	Support vector machine using biprofile Bayes, KNN features, k-tuple nucleotide compositions, and dinucleotide-based auto-covariance as features	Positive: promoters from Regulon DB.	24, 28, 32, 38, 54, 70	Web server and stand-alone http://flagshipnt.erc.monash.edu/MULTiPly/	2019	41	30
		[−60, 20].
iPromoter-2L	Multiwindow-based pseudo k-tuple nucleotide composition with physicochemical properties as features and Random Forest as a predictor	Positive: promoters from Regulon DB.	24, 28, 32. 38, 54, 70	Web server http://bioinformatics.hitsz.edu.cn/iPromoter-2L/	2018	40	180
		[−60, 20].
		Negative: randomly extracted from the middle regions of long coding sequences and convergent intergenic region (none of the promoters in each set has more than 0.8 pairwise sequence identity)

^aPositive, positive sequences, sequences expected to be promoters. Negative, negative sequences, sequences expected to not include promoters. The interval of the sequence with the boundary numbers related to a TSS is indicated within brackets ([−60, +20], [−60, +19], or [−200, +51]).

^bCitations checked on 3 May 2020.