**4. Tools for differential expression analysis of non-coding RNA**

**Tools Types Main Features/web link References**

sequences from Bacteria, Archaea, and Eukaryotes. http://evolveathome.com/

novel circRNAs, (ii) integrated miRNAtarget networks, (iii) expression profiles of circRNA isoforms, (iv) genomic annotations of circRNA isoforms, and (v) sequences of circRNA isoforms.

predicting circRNAs from those of noncircularized, expressed exons based on conformational and thermodynamic properties in the flanking introns. https://sourceforge.net/projects/

http://reprod.njmu.edu.cn/circrnadb

xypan1232/PredcircRNA

to predict circRNA. https://github.com/

and novel circRNAs in sequence data.

association of circular RNAs with diseases in human. http://gyanxet-beta.

(RNA Binding Proteins (RBP)- and miRNA-binding sites on human circRNAs. Allows to (i) identify potential circRNAs which can act as RBP sponges, (ii) design junctionspanning primers for specific detection of circRNAs of interest, (iii) design siRNAs for circRNA silencing, and (iv) identify potential internal ribosomal entry sites. https://circinteractome.nia.

species and 623 tRNA sequences from 104 species, provides various services such as graphical representations of tRNA secondary structures. trnadb.

[162]

[163]

[158]

[164]

[159]

[165]

[166]

[167]

[168]

snoRNA/snoRNA.php

circnet.mbc.nctu.edu.tw

predicircrnatool

circbase.org

com/circdb

nih.gov

bioinf.uni-leipzig.de

snoRNA Database Contains over 1000 snoRNA

120 Applications of RNA-Seq and Omics Strategies - From Microorganisms to Human Health

CircNet Provides the following resources: (i)

PredicircRNATool circRNA prediction Uses a machine learning method for

circRNADb circRNA database Contains 32,914 human circular RNAs.

PredcircRNA cirRNA prediction Applies a machine learning approach

CirsBase Database Provides scripts to identify known

Circ2Traits Database Contains a database of potential

CircInteractome Database Provides a web tool for mapping

tRNAdb Database Contains 12,000 tRNA genes from 577

**Table 5.** Overview of tools and databases for sequence analysis of other small ncRNAs.

Various tools allow for the detection of genes (mRNA or ncRNA) differentially expressed (DE) between two or more conditions or states from sequence data. The major differences among tools are their implemented statistical methods, input and output file formats as well as filtering steps for DE analyses. Many tools such as DESeq [169], edgeR [170], NBPSeq [171], TSPM [172], baySeq [173], EBSeq [174], NOISeq [175], SAMseq [176] and ShrinkSeq [177] use count data as input file, while others like limma [178] and Cufflinks use transformed data or BAM files (the binary version of sequence alignment data) as input, respectively. Tools that use count data can be divided in to two groups; parametric (DESeq [169], edgeR [170], NBPSeq [171], TSPM [172], baySeq [173], EBSeq [174]) and non-parametric methods (NOISeq [175], SAMseq [176]). For parametric methods, most softwares (baySeq [173], DESeq [169], NBPSeq [171], edgeR [170], EBSeq [174] and NBPSeq) use a negative binomial model to account for over dispersion except ShrinkSeq which has two options for distribution, either negative binomial or a zero-inflated negative binomial distribution. These methods also implement different statistical test approaches; DESeq, edgeR and NBPSeq perform a classical hypothesis testing approach while baySeq, EBSeq and ShrinkSeq apply Bayesian methods. The comparison of methods and performances have been done and reviewed by many authors [29, 179– 183]. In general, no single method performs well for all datasets. In a survey of performance of DE analyses methods, Conesa et al. [29] observed that limma package [178] performed well under many conditions. Many studies observed similar performances by DESeq and edgeR in ranking genes [29, 179–183]. However, DESeq is more conservative while edgeR is more liberal in controlling false discovery rate (FDR) [29]. Other tools such as SAMseq is better in controlling FDR while NOISeq is efficient in avoiding false positives [29].
