DNA sequences

Balance data

Group	Type	Literature	Year	Data	Acc		URL
Function Unequal length sequence	DNA Enhancers	iEnhancer-RF: Identifying enhancers and their strength by enhanced feature representation using random forest	2021	Download ↓ Training dataset: 742 strong enhancers 742 weak enhancers 1484 non-enhancers Testing dataset: 100 strong enhancers 100 weak enhancers 200 non-enhancers	5-CV layerI: 76.18% layerII: 62.53%	Independent testing layerI: 79.75% layerII: 85.00%	iEnhancer-RF
		iEnhancer-XG: interpretable sequence-based enhancers and their strength predictor	2021		10-CV layerI: 81.10% layerII: 66.74%	Independent testing layerI: 75.75% layerII: 63.50%	iEnhancer-XG
		iEnhancer-EL: identifying enhancers and their strength with ensemble learning approach	2018		jackknife test layerI: 78.03% layerII: 65.03%	Independent testing layerI: 74.75% layerII: 61.00%	iEnhancer-EL
		EnhancerPred: a predictor for discovering enhancers based on the combination and selection of multiple features	2016	Download ↓ Training dataset: 742 strong enhancers 742 weak enhancers 1484 non-enhancers	jackknife test layerI: 77.39%; layerII: 68.19%		EnhancerPred
		iEnhancer-2L: a two-layer predictor for identifying enhancers and their strength by pseudo k-tuple nucleotide composition	2015		jackknife test layerI: 76.89%; layerII: 61.93%		iEnhancer-2L
	Promoters	iPSW(2L)-PseKNC: A two-layer predictor for identifying promoters and their strength by hybrid features via pseudo K-tuple nucleotide composition	2019	Download ↓ Training dataset: promoter sequences: 3382 non-promoters: 3382	5-CV 84.06%		iPSW(2L)-PseKNC
	Nucleosome	NucPosPred: Predicting species-specific genomic nucleosome positioning via four different modes of general PseKNC	2018	Download ↓ Training dataset: S1: C. elegans 2567 nucleosome-forming 2608 nucleosome-inhibiting S2: D. melanogaster 2900 nucleosome-forming 2850 nucleosome-inhibiting	jackknife test S1: 92.29% S2: 88.26%		NucPosPred
	DNA sequence	DeepATT: a hybrid category attention neural network for identifying functional effects of DNA sequences	2021	Large Data Download ↓ Training dataset: 4400000 sequences; Validation dataset: 8000 sequences; Test dataset: 455024 sequences	AVAUROC: 0.94519 AVAUPR: 0.39522		DeepATT
Modification Equal length sequence	N⁴-methycytosine (4mC)	Identifying DNA N⁴-methylcytosine sites in the rosaceae genome with a deep learning model relying on distributed feature representation	2021	Download ↓ F. vesca: Training dataset: P-3457/N-3457; Testing dataset: P-864/N-864,4320,12960; R. chinensis: Training dataset: P-1938/N-1938; Testing dataset: P-483/N-483,2415,7245;	5-CV F.vesca: 0.8697 R.chinensis: 0.8541	Independent testing F.vesca: P/N=1:1 0.8632; P/N=1:5 0.8632; P/N=1:15 0.8412 R.chinensis: P/N=1:1 0.8490; P/N=1:5 0.8400; P/N=1:15 0.8477	4mC-w2vec
		Escherichia Coli DNA N⁴-Methycytosine Site Prediction Accuracy Improved by Light Gradient Boosting Machine Feature Selection Technology	2020	Download ↓ E.coli: Training dataset: P-388 /N-388 Testing dataset: P-134 /N-134	10-CV Training set: 85.4%	Independent testing 83.2%	EC4mC-SVM
		Developing a Multi-Layer Deep Learning Based Predictive Model to Identify DNA N⁴-Methylcytosine Modifications	2020	Large Data Download ↓ A. thaliana: P-20000/ N-20000; C. elegans: P-20000/ N-20000; D. melanogaster: P-20000/ N-20000;	10-CV A. thaliana: 84.4%; C. elegans: 89.3%; D. melanogaster: 87.1%		Deep4mcPred
		Iterative feature representations improve N⁴-methylcytosine site prediction	2019	Download ↓ DatasetI: C. elegans: P-1554 /N-1554; D. melanogaster: P-1769 /N-1769; A. thaliana: P-1978 /N-1978; E. coli: P-388/N-388; G. subterraneus: P-905/N-905; G. pickeringii: P-569/N-569;	Cross_validation C. elegans: 0.880; D. melanogaster: 0.874; A. thaliana: 0.825; E. coli: 0.894; G. subterraneus: 0.886;G. pickeringii: 0.907		4mcPred-IFL
		Exploring sequence-based features for the improved prediction of DNA N⁴-methylcytosine sites in multiple species	2019		10-CV C.elegans: 0.815; D.melanogaster: 0.830; A.thaliana: 0.787; E.coli: 0.833; G.subterruneus: 0.837; G.pickeringii: 0.860		4mcPred-SVM
		4mCPred: machine learning methods for DNA N⁴-methylcytosine sites prediction	2019		jackknife test C.elegans: 87.71%; D.melanogaster: 87.79%; A.thaliana: 83.37%; E.coli: 94.97%; G.subterruneus: 91.04%; G.pickeringii: 90.89%	Independent testing C.elegans: 82.21%; D.melanogaster: 82.63%; A.thaliana: 76.52%; E.coli: 82.69%; G.subterruneus: 83.33%; G.pickeringii: 77.63%	4mCPred
		Meta-4mCpred: A Sequence-Based Meta-Predictor for Accurate DNA 4mC Site Prediction Using Effective Feature Representation	2019	Download ↓ Training dataset: DatasetI Testing dataset:C. elegans: P-750/ N-750; D. melanogaster: P-1000/ N-1000; A. thaliana: P-1250/ N-1250; E. coli: P-134/N-134; G. subterraneus: P-350/N-350; G. pickeringii: P-200/ N-200;	10-CV C.elegans: 0.826; D.melanogaster: 0.842; A.thaliana: 0.792; E.coli: 0.848; G.subterruneus: 0.855; G.pickeringii: 0.891	Independent testing C.elegans: 0.870; D.melanogaster: 0.906; A.thaliana: 0.855; E.coli: 0.825; G.subterruneus: 0.850; G.pickeringii: 0.850	Meta-4mCpred
		4mCpred-EL: An Ensemble Learning Framework for Identification of DNA N⁴-Methylcytosine Sites in the Mouse Genome	2019	Download ↓ mouse genome Training dataset: P-800/ N-800 Testing dataset: P-180/ N-180	10-CV Training dataset: 0.795±0.001	Independent testing 0.798±0.011	4mCpred-EL
	N⁶-methyladenine (6mA)	MGF6mARice: prediction of DNA N6-methyladenine sites in rice by exploiting molecular graph feature and residual block	2022	Large Data Download ↓ Benchmark datasets Rice-Chen: P-154000/ N-154000; Rice-Lv: P-880/ N-880; Imbalanced datasets Rice-Chen(1:5): P-176/ N-880; Rice-Chen(1:10): P-88/ N-880; Rice-Chen(1:20): P-44/ N-880; Rice-Lv(1:5): P-30800/ N-154000; Rice-Lv(1:10): P-15400/ N-154000; Rice-Lv(1:20): P-7700/ N-154000; Independent datasets NIP_10000: P-10000/ N-10000; A.thaliana: P-15937/ N-15937; D.melanogaster: P-11191/ N-11191; R.chinensis: P-11815/ N-11815;	10-CV Rice-Chen: 97.00% Rice-Lv: 96.00%	Independent testing NIP_10000: 88.00%; A.thaliana: 77.00%; D.melanogaster:82.00%; R.chinensis:86.00%;	MGF6mARice
		Deep6mA: A deep learning framework for exploring similar patterns in DNA N6-methyladenine sites across different species	2021	Large Data Download ↓ Rice: P-154000/ N-154000; A.thaliana: P-98483/ N-98483; F.vesca: P-1417/ N-1417; R.chinensis: P-5733/ N-5733;	5-CV Rice: ACC 94.01%; A.thaliana: AUC 0.954; F.vesca: AUC 0.982; R.chinensis: AUC 0.961;		Deep6mA
		A Convolutional Neural Network Using Dinucleotide One-hot Encoder for identifying DNA N⁶-Methyladenine Sites in the Rice Genome	2021	Large Data Download ↓ Rice Genome Dataset-I: P-154000/ N-154000 Dataset-II: P-880/ N-880	5-CV Dataset-I: 93.82%	Independent testing Dataset-II: 96.19%	iRicem6A-CNN
		6mA-RicePred: A Method for Identifying DNA N⁶-Methyladenine Sites in the Rice Genome Based on Feature Fusion	2020		10-CV Dataset-II: 87.27%	Independent testing Dataset-I: 85.65%	6mA-RicePred
		iIM-CNN: Intelligent Identifier of 6mA Sites on Different Species by Using Convolution Neural Network	2019	Download ↓ cross-species: P-2768/ N-2716; Rice: P-880/ N-880; M. musculus: P-1934/ N-1934;	5-CV cross-species: 0.824; Rice: 0.875; M.musculus: 0.969		iIM-CNN
		csDMA: an improved bioinformatics tool for identifying DNA 6mA modifications via Chou's 5-step rule	2019	Download ↓ Cross-species Training dataset: P-2214/ N-2214; Testing dataset: P-554/ N-502; Rice: P-880/ N-880; M. musculus: P-1934/ N-1934;	5-CV Cross-species Training dataset: 0.799; Rice: 0.861; M.musculus: 0.966	Independent testing Cross-species: 0.813;	csDMA

Note: 5-CV：5-fold cross validation; 10-CV: 10-fold cross validation；P: Positive samples; N: Negative samples;

Imbalance data

Group	Type	Literature	Year	Data	Imbalance Algorithms	Acc		URL
Function Unequal length sequence	Promoters	iPro2L-PSTKNC: A Two-Layer Predictor for Discovering Various Types of Promoters by Position Specific of Nucleotide Composition	2021	Download ↓ promoter sequences: σ²⁴: 484; σ²⁸: 134; σ³²: 291; σ³⁸: 163; σ⁵⁴: 94; σ⁷⁰: 1694 non-promoter sequences: 2860	SMOTE	5-CV layerI: 90.05%	5-CV layerII: σ²⁴: 97.75%; σ²⁸: 99.84%; σ³²: 98.66%; σ³⁸: 99.06%; σ⁵⁴: 99.94%; σ⁷⁰: 94.19%	iPro2L-PSTKNC
Function Unequal length sequence	Promoters	iPromoter-2L: a two-layer predictor for identifying promoters and their types by multi-window-based PseKNC	2018		IHTS	5-CV layerI: 81.68%	5-CV layerII: σ²⁴: 93.50%; σ²⁸: 96.85%; σ³²: 94.41%; σ³⁸: 94.69%; σ⁵⁴: 94.04%; σ⁷⁰: 80.66%	iPromoter-2L
Modification Equal length sequence	5-methylcytosine (5mC)	iPromoter-5mC: A Novel Fusion Decision Predictor for the Identification of 5-Methylcytosine Sites in Genome-Wide DNA Promoters	2020	Large Data Download ↓ Training dataset: P-55800/ N-13950 Testing dataset: P-658861/ N-164715	DSM	5-CV Training dataset: 90.16%	Independent testing 90.22%	iPromoter-5mC
	5-methylcytosine (5mC)	iDNA-Methyl: Identifying DNA methylation sites via pseudo trinucleotide composition	2015	Download ↓ Methylation samples: 787 Non-methylation samples: 1639	NCR/SMOTE	jackknife test 77.49%		iDNA‐Methyl

Note: 3-CV：3-fold cross validation; 5-CV: 5-fold cross validation；P: Positive samples; N: Negative samples;
SMOTE: Synthetic Minority Oversampling Technique;
IHTS: Inserting Hypothetical Training Samples;
DSM: Down-Sampling Method;
NCR: Neighborhood Cleaning Rule;

Classification based on DNA sequence