RNA sequences

Balance data

Group	Type	Literature	Year	Data	Acc		URL
Function Unequal length sequence	eukaryotic mRNA	SubLocEP: a novel ensemble predictor of subcellular localization of eukaryotic mRNA based on machine learning	2021	Download ↓ Training dataset: Cytoplasm: 5310; Endoplasmic reticulum: 1185; Extracellular region: 710; Mitochondria: 350; Nucleus: 4855; Testing dataset1: Cytoplasm: 1066; Endoplasmic reticulum: 241; Extracellular region: 145; Mitochondria: 71; Nucleus: 976; Testing dataset2: Cytosol: 91; Nucleus: 148; Testing dataset3: Endoplasmic reticulum: 131; Nucleus: 131;	5-CV Training set 0.659±0.006	Independent testing1 0.601; Independent testing2 0.506; Independent testing3 0.37;	SubLocEP
	lncRNA	PlncRNA-HDeep: plant long noncoding RNA prediction using hybrid deep learning based on two encoding styles	2021	Download ↓ Dataset: Positive Dataset: 18000; Negative Dataset: 18000; (training set (80%) and test data (20%))	5-CV 96.5%		PlncRNA-HDeep
	lncRNA	PredLnc-GFStack: A Global Sequence Feature Based on a Stacked Ensemble Learning Method for Predicting lncRNAs from Transcripts	2019	Download ↓ Human: Main: P-35760/ N-20299; Independent: P-1500/ N-1500; Mouse: Main: P-23987/ N-11746; Independent: P-1500/ N-1500; CPPred: Human-Testing: P-8557/ N-8241; Mouse-Testing: P-31102/ N-19930; Zebrafish-Testing: P-15594/ N-10662; Fruit-fly-Testing: P-17400/ N-4098; S.cerevisiae-Testing: P-6713/ N-413; Integrate-Testing: P-13903/ N-13903;	10-CV Human-Main: 0.895; Mouse-Main: 0.914;	Independent testing Human-Main: Human-Testing: 0.968; Mouse-Testing: 0.941; Zebrafish-Testing: 0.901; Fruit-fly-Testing: 0.940; S.cerevisiae-Testing: 0.960; Integrate-Testing: 0.907; Mouse-Main: Human-Testing: 0.887; Mouse-Testing: 0.944; Zebrafish-Testing: 0.843; Fruit-fly-Testing: 0.917; S.cerevisiae-Testing: 0.942; Integrate-Testing: 0.871;	PredLnc-GFStack
	Non-Coding RNAs	ncRFP: A Novel end-to-end Method for Non-Coding RNAs Family Prediction Based on Deep Learning	2021	Download ↓ 6320 ncRNAs sequence [PS: One part of each family is randomly selected to form a test set and the rest to form a train set, so that all ncRNAs sequences can form 10-fold cross-validation train sets and test sets]	10-CV 0.7972		ncRFP
		ncRDense: A novel computational approach for classification of non-coding RNA family by deep learning	2021	Download ↓ Training dataset: 6320; Testing dataset: 2600;	10-CV 0.9519	Independent testing 0.9510	ncRDense
		ncRDeep: Non-coding RNA classification with convolutional neural network	2020		10-CV 88.04%		ncRDeep
	pre-miRNA	ASRmiRNA: Abiotic Stress-Responsive miRNA Prediction in Plants by Using Machine Learning Algorithms with Pseudo K-Tuple Nucleotide Compositional Features	2022	Download ↓ miRNA training dataset: P-376/ N-376; Pre-miRNA training dataset: P-251/ N-251; Pre-miRNA+miRNA training dataset: 251 stress-responsive Pre-miRNAs/ 251 non-stress-responsive Pre-miRNAs; 251 stress-responsive miRNAs/ 251 non-stress-responsive miRNAs; Independent dataset: 70 stress-responsive miRNAs/ 100 non-stress-responsive miRNA; 70 stress-responsive Pre-miRNAs/ 100 non-stress-responsive Pre-miRNA	5-CV miRNA: 65.33%; Pre-miRNA: 66.40%; miRNA + Pre-miRNA: 71.40%;	Independent testing miRNA: 62.33%; Pre-miRNA: 64.85%; miRNA + Pre-miRNA: 69.21%;	ASRmiRNA
	pre-miRNA	Deep neural networks for human microRNA precursor detection	2020	Download ↓ Training dataset: 2408 sequences; Validation dataset: 602 sequences; Testing dataset: 752 sequences;	10-CV CNN: 87.24±1.80%; RNN: 88.44±1.80%		DNNPreMiR
	circRNA	CirRNAPL: A web server for the identification of circRNA based on extreme learning machine	2020	Download ↓ circRNA vs PCG: Training dataset: circRNA-10000/ PCG-8000; Testing dataset: circRNA-4084/ PCG-1533; circRNA vs lncRNA: Training dataset: circRNA-10000/ lncRNA-10000; Testing dataset: circRNA-4084/ lncRNA-9722; Stem cell vs not: Training dataset: circRNA-1800/non-circRNA-1800; Testing dataset: circRNA-282/non-circRNA-282;	10-CV circRNA vs PCG: 0.815; circRNA vs lncRNA: 0.802; Stem cell vs not: 0.782;	Independent testing circRNA vs PCG: 0.827; circRNA vs lncRNA: 0.854; Stem cell vs not: 0.812;	CirRNAPL
	tRNA	Improving tRNAscan-SE Annotation Results via Ensemble Classifiers	2015	Download ↓ Training dataset: Positive tRNA sequences: 623 Negative tRNA sequences: 1183	10-CV 95.1%%		tRNA-Predict^∗
	piRNA	Identification of Functional piRNAs Using a Convolutional Neural Network	2020	Download ↓ Training dataset: First Layer: 709/709 Second Layer:1418/1418	5-CV First Layer: 93.59%; Second Layer: 90.13%;		2S-piRCNN
	piRNA	2L-piRNA: A Two-Layer Ensemble Classifier for Identifying Piwi-Interacting RNAs and Their Function	2017		5-CV First Layer: 86.1%; Second Layer: 77.6%;		2L-piRNA
Modification Equal length sequence	N1-methyladenosine (m1A) 5-methylcytosine (m5C) N6-methyladenosine (m6A) Pseudouridine Adenosine-to-inosine (A-to-I)	iMRM: a platform for simultaneously identifying multiple kinds of RNA modifications	2020	Download ↓ *H.sapiens:* m1A: P-6366/ N-6366; m5C: P-120/ N-120; m6A: P-1130/ N-1130; Pseudouridine: P-495/ N-195; A-to-I: P-3000/N-3000; *S.cerevisiae:* m1A: P-483/ N-483; m5C: P-211/ N-211; m6A: P-1307/ N-1307; Pseudouridine: P-313/ N-314; *M.musculus:* m1A: P-1064/ N-1064; m5C: P-97/ N-97; m6A: P-725/ N-725; Pseudouridine: P-472/ N-472;	10-CV *H.sapiens:* m1A: 99.47%; m5C: 93.33%; m6A: 91.28%; Pseudouridine: 66.47%; A-to-I: 91.73%; *S.cerevisiae:* m1A: 98.87%; m5C: 100%; m6A: 78.41%; Pseudouridine: 71.91%; *M.musculus:* m1A: 99.29%; m5C: 99.00%; m6A: 89.59%; Pseudouridine: 74.48%;		iMRM
	Pseudouridine	RF-PseU: A Random Forest Predictor for RNA Pseudouridine Sites	2020	Download ↓ H. sapiens: Training dataset: P-495/N-495; Testing dataset: P-100/ N-100; S. cerevisiae: Training dataset: P-314/N-314; Testing dataset: P-100/ N-100; M. musculus: Training dataset: P-472/N-472;	10-CV H. sapiens: 64.3%; S. cerevisiae: 74.8%; M. musculus:74.8%;	Independent testing H. sapiens: 75.0% S. cerevisiae: 77.0%	RF-PseU
		iPseU-CNN: Identifying RNA Pseudouridine Sites Using Convolutional Neural Networks	2019		5-CV H. sapiens: 66.68%; S. cerevisiae: 68.15%; M. musculus:71.81%;	Independent testing H. sapiens: 69.00% S. cerevisiae: 73.50%	/
		Is There Any Sequence Feature in the RNA Pseudouridine Modification Prediction Problem?	2020	Download ↓ H. sapiens: Training dataset: P-495/N-495; Testing dataset: P-100/ N-100; S. cerevisiae: Training dataset: P-319/N-319; Testing dataset: P-100/ N-100; M. musculus: Training dataset: P-495/ N-495;	5-CV H. sapiens: 62.73%; S. cerevisiae: 70.54%; M. musculus: 71.72%;	Independent testing H. sapiens: 60.20% S. cerevisiae: 77.0%	/
	5-methylcytosine (m5C)	Prediction of m5C Modifications in RNA Sequences by Combining Multiple Sequence Features	2020	Download ↓ A. thaliana: Training dataset: P-5289 /N-5289 Testing dataset: P-1000/ N-1000	10-CV A. thaliana: 73.06%	Independent testing A. thaliana: 80.15%	/
	5-methylcytosine (m5C)	Evaluation of different computational methods on 5-methylcytosine sites identification	2020	Download ↓ H. sapiens: P-120/ N-120; M. musculus: P-97/ N-97; S. cerevisiae: P-211/ N-211; A. thaliana: Training dataset: P-5289/ N-5289; Testing dataset: P-1000/N-1000;	jackknife test H. sapiens: 90.8%; M. musculus: 100%; S. cerevisiae: 100%; 10-CV A.thaliana: 70.7%;	Independent testing A.thaliana: 74.0%	iRNA-m5C
	5-hydroxymethylcytosine (5hmC)	Prediction of RNA 5-Hydroxymethylcytosine Modifications Using Deep Learning	2021	Download ↓ Training dataset: P-662/ N-662	5-CV 0.81		iRhm5CNN
	5-hydroxymethylcytosine (5hmC)	iRNA5hmC: the First Predictor to Identify RNA 5-Hydroxymethylcytosine Modifications Using Machine Learning	2020	Download ↓ Training dataset: P-662/ N-662	5-CV 65.48%		iRNA5hmC
	N6-methyladenosine (m6A)	DNN-m6A: A Cross-Species Method for Identifying RNA N6-methyladenosine Sites Based on Deep Neural Network with Multi-Information Fusion	2021	Download ↓ Human Brain: Training dataset: P-4605/ N-4605; Testing dataset: P-4604/4604; Kindey: Training dataset: P-4574/ N-4574; Testing dataset: P-4573/ N-4573; Liver: Training dataset: P-2634/ N-2634; Testing dataset: P-2634/ N-2634; Mouse Brain: Training dataset: P-8025/ N-8025; Testing dataset: P-8025/ N-8025; Heart: Training dataset: P-2201/ N-2201; Testing dataset: P-2200/ N-2200; Kidney: Training dataset: P-3953/ N-3953; Testing dataset: P-3952/ N-3952; Liver: Training dataset: P-4133/ N-4133; Testing dataset: P-4133/ N-4133; Testis: Training dataset: P-4704/ N-4704; Testing dataset: P-4706/ N-4706; Rat Brain: Training dataset: P-2352/ N-2352; Testing dataset: P-2351/ N-2351; Kidney: Training dataset: P-3433/ N-3433; Testing dataset: P-3432/ N-3432; Liver: Training dataset: P-1762/ N-1762; Testing dataset: P-1762/ N-1762;	5-CV Human: Brian: 0.7378; Kidney: 0.8048; Liver: 0.8130; Mouse: Brian: 0.7936; Heart: 0.7617; Kidney: 0.8196; Liver: 0.7358; Testis: 0.7662; Rat Brain: 0.7827; Kidney: 0.8338; Liver: 0.8263;	Independent testing Human: Brian: 0.7327; Kidney: 0.7989; Liver: 0.8096; Mouse: Brian: 0.7859; Heart: 0.511; Kidney: 0.8087; Liver: 0.7295; Testis: 0.7712; Rat Brain: 0.7799; Kidney: 0.8304; Liver: 0.8164;	DNN-m6A
		Computational identification of N6-methyladenosine sites in multiple tissues of mammals	2020		5-CV Human: Brian: 71.26%; Kidney: 78.99%; Liver: 80.13%; Mouse: Brian: 78.75%; Heart: 72.79%; Kidney: 79.98%; Liver: 70.59%; Testis: 74.40%; Rat Brain: 75.96%; Kidney: 81.78%; Liver: 80.90%;	Independent testing Human: Brian: 71.1%; Kidney: 77.76%; Liver: 79.01%; Mouse: Brian: 78.26%; Heart: 71.3%; Kidney: 79.31%; Liver: 68.79%; Testis: 73.54%; Rat Brain: 75.14%; Kidney: 81.42%; Liver: 79.85%;	iRNA-m6A
		HSM6AP: a high-precision predictor for the Homo sapiens N6-methyladenosine (m^6A) based on multiple weights and feature stitching	2021	Large Data Download ↓ H.sapiens: Training dataset: P-14025/N-14025; Independent dataset1: P-23478/ N-23478; Independent dataset2: P-40742/ N-40742; Independent dataset3: P-15696/ N-15696;	5-CV Full transcript: A549: 0.973; CD8T: 0.968; HEK293-abacm: 0.978; HEK293_sysy: 0.985; Hela: 0.984; MOLM13: 0.968; Mature: A549: 0.908; CD8T: 0.89; HEK293-abacm: 0.97; HEK293_sysy: 0.899; Hela: 0.901; MOLM13: 0.967;	Independent testing1 Full transcript: AUC: 0.976; Mature mRNA: AUC: 0.899; Independent testing2 Full transcript: AUC: 0.981; Mature mRNA: AUC: 0.914; Independent testing3 Full transcript: AUC: 0.96776; Mature mRNA: AUC: 0.890;	HSM6AP
		WHISTLE: a high-accuracy map of the human N6-methyladenosine (m6A) epitranscriptome predicted using a machine learning approach	2019	Download ↓ A549: P-23478/N-234780; CD8T: P-19677/N-196770; Hela: P-37183/N-371830; HEK293(sys): P-12051/N-120510; HEK293(abcm): P-9536/N-95360;	5-CV Full transcript: AUC (A549: 0.977; CD8T: 0.976; Hela: 0.976; HEK293(sys): 0.976; HEK293(abacm): 0.975; MOLM13: 0.979;) Mature mRNA: AUC (A549: 0.938; CD8T: 0.940; Hela: 0.936; HEK293(sysy): 0.941; HEK293(abacm): 0.942; MOLM13: 0.943);	Independent Testing Full Transcript: AUC (A549: 0.973; CD8T: 0.947; Hela: 0.962; HEK293(sysy): 0.954; HEK293(abacm): 0.976; MOLM13: 0.947); Mature mRAN: AUC (A549: 0.923; CD8T: 0.922; Hela: 0.911; HEK293(sysy): 0.904; HEK293(abacm): 0.857; MOLM13: 0.856);	WHISTLE
		iMethyl-Deep: N6-Methyladenosine Identification of Yeast Genome with Automatic Feature Extraction Technique by Using Deep Learning Algorithmn	2020	Download ↓ S.cerevisiae: Dataset-1-M6A2614: P-1307/ N-1307 Dataset-2-M6A6540: P-3270/ N-3270	10-CV Dataset-1-M6A2614: 89.19%; Dataset-2-M6A6540: 87.44%;		iMethyl-Deep
		Integration of deep feature representations and handcrafted features to improve the prediction of N6-methyladenosine sites	2018		10-CV Dataset-1-M6A2614: 80.50%		DeepM6APred
		Detecting N6-methyladenosine sites from RNA transcriptomes using ensemble Support Vector Machines	2016		jackknife test Dataset-1-M6A2614: 78.35%		RAM-ESVM
		HLMethy: a machine learning-based model to identify the hidden labels of m6A candidates	2019	Large Data Download ↓ Training dataset: P-11799/ N-11799 Testing dataset: P-2976/ N-2976	5-CV Training dataset: 0.726	Independent Testing 0.727	HLMethy
		Gene2vec: gene subsequence embedding for prediction of mammalian N6-methyladenosine sites from mRNA	2018	Large Data Download ↓ Training dataset: 88579(P:N=1:1) Testing dataset: 88227(P:N=1:10)	10-CV H. sapiens: AUC 0.8414; M. musculus: AUC 0.8145;	Independent testing AUC 0.8414	Gene2vec
		BERMP: a cross-species classifier for predicting m6A sites by integrating a deep learning algorithm and a random forest approach	2018	Download ↓ S. cerevisiae: Training dataset: P-1307/ N-1307; Testing dataset: P-207/ N-207; A. thaliana: Training dataset: P-2100/ N-2100; Testing dataset: P-418/ N-418; Mammalian: Training dataset: P-40000/ N-400000; Testing dataset: P-10000/ N-100000;	10-CV S. cerevisiae: 71.26%; 5-CV A. thaliana: 85.95%; Mammalia: Full transcript: 87.80%; Mature mRNA: 86.14%;	Independent testing S. cerevisiae: 68.84%; A. thaliana: 87.20%; Mammalia: Full transcript: 88.34%; Mature mRNA:86.32%;	BERMP
	2’-O-methylation	NmRF: identification of multispecies RNA 2’-O-methylation modification sites from RNA sequences	2022	Download ↓ H. sapiens: P-261/ N-329; S.cerevisiae: P-89/ N-189; M.musculus: P-10/ N-35;	10-CV H. sapiens:89.069%; S.cerevisiae:93.885%;	Independent testing H. sapiens: 86.88%	NmRF
		iRNA-PseKNC(2methyl): Identify RNA 2’-O-methylation sites by convolution neural network and Chou’s pseudo components	2019	Download ↓ H. sapiens: Training dataset: P-147/ N-147	5-CV 98.27%		/
		iRNA-2OM: A Sequence-Based Predictor for Identifying 2'-O-Methylation Sites in Homo sapiens	2018		5-CV 97.95%		iRNA-2OM
	N7-methylguanosine (m7G)	m7G-DPP: Identifying N7-methylguanosine sites based on dinucleotide physicochemical properties of RNA	2021	Download ↓ Training dataset: P-741/ N-741	10-CV 95.42%	jackknife test 95.55%	m7G-DPP
		Iterative feature representation algorithm to improve the predictive performance of N7-methylguanosine sites	2020		10-CV 92.50%		m7G-IFL
		iRNA-m7G: Identifying N7-methylguanosine Sites by Fusing Multiple Features	2019		jackknife test 89.88%		iRNAm7G
		m7GPredictor: An improved machine learning-based model for predicting internal m7G modifications using sequence properties	2020	Download ↓ Training dataset: P-595/ N-595 Testing dataset: P-150/ N-150	10-CV 86.00%	Independent testing 86.00%	m7GPredictor

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	sgRNA	SgRNA-RF: identification of SgRNA on-target activity with imbalanced datasets	2021	Download ↓ G17 sequences: Training dataset: P-830/ N-900; Testing dataset: P-229/ N-3351; Gr sequences: Training dataset: P-550/ N-320; Testing dataset: P-181/ N-118; Gnr sequences: Training dataset: P-536/ N-180; Testing dataset: P-135/ N-57; Gm sequences: Training dataset: P-664/ N-180; Testing dataset: P-166/ N-51; G5 sequences: hek293t: Training dataset: P-1615/ N-428; Testing dataset: P-404/ N-108; hct116: Training dataset: P-3090/ N-428; Testing dataset: P-783/ N-108; hela: Training dataset: P-5923/ N-428; Testing dataset: P-782/ N-108; h160: Training dataset: P-1973/ N-428; Testing dataset: P-494/ N-108;	CS-Smote	10-CV G17: 84.7%; Gr: 69.4%; G5: hct116: 96.9%; hek293t: 94.0%; hela: 97.74%; h160: 94.10%;	Independent testing G17: 86.3%; Gr: 91.6%; Gnr: 89.4%; Gm: 93.8%; G5: hct116: 96.50%; hek293t: 78.7%; hela: 97.9%; h160: 97.3%;	SgRNA-RF
Modification Equal length sequence	N2-methylguanosine (m2G)	RFhy-m2G: Identification of RNA N2-methylguanosine modification sites based on random forest and hybrid features	2021	Download ↓ Training set: H. sapiens: P-41/ N-541; M. musculus: P-27/ N-427; S. cerevisiae: P-60/ N-283; Testing dataset: H. sapiens: P-5/ N-60; M. musculus: P-3/ N-47; S. cerevisiae: P-7/ N-31;	SMOTE	5-CV H. sapiens: 0.9982; M. musculus: 1.0; S. cerevisiae: 0.9965;	Independent testing H. sapiens: 0.9417; M. musculus: 0.9043; S. cerevisiae: 1.0;	/
	N6-methyladenosine (m6A) Pseudouridine N1-methyladenosine (m1A) N6,2′-O-dimethyladenosine (m6Am) 5-methylcytosine (m5C) 2′-O-methyladenosine (Am) 2′-O-methylcytidine (Cm) 2′-O-methylguanosine (Gm) 2′-O-methyluridine (Um) N7-methylguanosine (m7G) 5-Methyluridine (m5U) Inosine (I)	Attention-based multi-label neural networks for integrated prediction and interpretation of twelve widely occurring RNA modifications	2021	Large Data Download ↓ m6A: 65178; Pseudouridine: 3137; m1A: 16380; m6Am: 2447; Am: 1591; Cm: 1878; Gm: 1471; Um: 2253; m5C: 12936; m7G: 1036; m5U: 1696; I: 52618;	OHEM and UW	Am: 78.00%; Cm: 82.00%; Gm: 89.00%; Um: 82.00%; m1A: 72.00%; m5C: 85.00%; m5U: 92.00%; m6A: 80.00%; m6Am: 83.00%; m7G: 65.00%; Pseudouridine: 84.00%; I: 70.00%;		MultiRM
	Dihydrouridine (D)	Accurate identification of RNA D modification using multiple features	2021	Download ↓ Multi-species Training dataset: P-140/ N-298 Testing dataset: P-36/ N-76	SMOTEEEN	jackknife test Training dataset: 97.24%	Independent testing 93.75%	/
	5-methylcytosine (m5C)	iRNA-m5C_NB: A Novel Predictor to Identify RNA 5-Methylcytosine Sites Based on the Naive Bayes Classifier	2020	Download ↓ H. sapiens: Met240: Training dataset: P-120/ N-120; Met935: Training dataset: P-127/ N-808; Testing dataset: Test1157: P-157/ N-1000;	SMOTEENN	jackknife test Met935: 82.20%;	Independent testing Test1157: 74.85%	/

Note: 5-CV：5-fold cross validation; 10-CV: 10-fold cross validation；P: Positive samples; N: Negative samples;
SMOTE: Synthetic Minority Oversampling Technique;
IHTS: Inserting Hypothetical Training Samples;
SMOTEENN: Synthetic Minority Over-sampling Technique (SMOTE) and undersampling method Edited Nearest Neighbours (ENN);
OHEM: online hard examples mining;
UW: uncertain weights;

Classification based on RNA sequence