Table 1.

State-of-the-art computational intelligence (CI) techniques for finding non-coding RNA (ncRNA) genes from 2001 to 2016.

Cited	Approach	4 CMP	Results (%)	Methodology and Online Resource Tools
[13]	A computational approach to identify genes for functional RNAs in genomic sequences.	√	2 S: 90%, 3 P: 99%	NN and SVM. Online tool unavailable.
[14]	To detect ncRNA sequences.	×	−−−−−−	The support vector machine (SVM) algorithm was implemented in graphical processing units (GPUs) based parallel technology. Online tool unavailable.
[15]	To differentiate between well-known classes and target predicted classes of messenger RNA (mRNA).	√	−−−−−−	A new web-based interface was developed to detect ncRNAs. Available at http://biotools.ceid.upatras.gr/ncrnaclass/.
[16]	To identify ncRNA a positive sample only learning algorithm is introduced.	×	1 A: 80%	The SVM used as the core learning machine assessed by 5-fold-validation in recovery of known ncRNA. Data available online at (http://bioinformatics.oxfordjournals.org/content/22/21/2590/suppl/DC1)
[17]	To introduce a method to differentiate between coding or non-coding RNA.	×	3 P: 97%, 2 S: 98%	Supervised machine learning SVM is used to classify transcripts according to features they would have if transcripts coded for proteins. Online data source of mRNA at: RNAdb (http://research.imb.uq.edu.au/rnadb).
[20]	To identify ncRNA using six features extracted from transcript’s nucleotide sequence.	×	−−−−−−	SVM (coding potential calculator ((CPC)) to identify ncRNA using six features extracted from transcript’s nucleotide sequence. Dataset used Rfam and RNAdb for noncoding and EMBL CDS for coding. Online web-based interface available of CPC at http://cpc.cbi.pku.edu.cn.
[23]	The prediction of ncRNA genes using boosted genetic programming.	×	1 A: 80%	The GA and 10-fold cross validation was used to train and test the learning machine. Online tool unavailable.
[25]	To classify micro RNAs (miRNAs) and to differentiate between normal and tumor tissues.	√	−−−−−−	A multi-objective algorithm was developed by using four classifiers such as random tree (RT), random forest (RF), sequential minimal optimization (SMO) and logistic regression (LR).
[26]	To automatically predict miRNA target.	√	F-measure: 0.95	The deep neural-network (DNN) was utilized to increase F-measure by 25% for prediction of miRNA targets. Available at (http://data.snu.ac.kr/pub/deepTarget)
[27]	To predict miRNAs targets.	×	1 A: 90%, 2 S: 88%, 3 P: 94%	Contrast relaxing and convolutional neural network (CNN) methods. Online tool unavailable.
[28]	To predict new miRNA, known as pre-miRNAs.	×	1 A: 99.9%, 2 S: 99.8%, 3 P: 100%	A neural networks (NNs) classifier was used to predict miRNA. Online tool unavailable.
[29]	To improve the performance and to predict the regulation of miRNA.	×	−−−−−−−−	The authors utilized a NNs classifier to predict miRNA. Online tool unavailable.
[30]	To predict a real pre-miRNA or a pseudo pre-miRNA.	√	1 S: 97.40%, 2 P: 95.85%	The authors utilized a multilayer artificial neural network (ANN) classifier. Online tool unavailable.
[31]	A de novo prediction algorithm to identify ncRNA using features derived from sequence and structure of known ncRNA.	×	2 S: 68%, 3 P: 70%, 1 A: 70%	NN-based meta-learner de novo predictor using folding, ensemble, and structure-based features. Online data and program found at: http://csbl.bmb.uga.edu/publications/materials/tran/
[32]	The 15 disease related ncRNAs sequences are utilized from the ncRNAs with Alzheimer disease.	×	−−−−−−	From the NONCODE database [19], 15 disease related ncRNA sequences were selected for mapping and comparison. The ncRNA sequences in the cellular process and the base content in these sequences have almost the same Z-curves even though they are coming from different organisms. Online tool unavailable.
[33]	To identify ncRNA genes using a genetic algorithm (GA).	×	−−−−−−	The observed sequence in real sequence data is used to motivate the use of GAs to quickly reject regions of the search space of ncRNAs. Online tool unavailable.
[34]	To identify ncRNA using covariance searching.	×	−−−−−−	The covariance models for ncRNA gene finding is extremely powerful and also extremely computationally demanding. Online tool unavailable.
[35]	A comparative genomic approach is used to detect ncRNA.	×	−−−−−−	Developed an efficient clustering method for finding potential ncRNAs in bacteria by clustering genomic sequences. Online tool unavailable.
[36]	To identify real and pseudo miRNA using SVM with features that are present in local structure-sequence.	×	1 A: 90%	A method to classify real and pseudo miRNA by applying SVM using local structure sequence features. Online tool unavailable.
[37]	Computational identification of ncRNAs in Saccharomyces cerevisiae by comparative genomics.	×	−−−−−−	Computational screen followed by Northern blot and transcript sequencing. Online tool unavailable. Data set is available only at: http://genome.cshlp.org/content/13/6b/1301/suppl/DC1.
[38]	Identification of putative noncoding RNAs among the RIKEN mouse full-length cDNA collection.	×	−−−−−−	The authors identified nine ncRNAs. Online tool unavailable. Data set is available only at: http://genome.cshlp.org/content/13/6b/1301/suppl/DC1.
[39]	The 19 candidate ncRNAs were identified including one with significant homology.	×	−−−−−−	The author used base-composition statistics method to find variety of ncRNAs. Online tool unavailable.
[40]	ncRNA gene detection using comparative sequence analysis.	√	2 S: 97.3%, 3 P: 100%	Comparative sequence analysis algorithm with “pair grammars” based on stochastic and hidden Markov models (HMM). Online tool unavailable.

¹ A: Accuracy, ² S: Sensitivity, ³ P: Specificity, and ⁴ CMP: Comparisons, √: Compared and ×: Not compared.