C-mii: a tool for plant miRNA and target identification
© Numnark et al.; licensee BioMed Central Ltd. 2012
Published: 13 December 2012
Skip to main content
© Numnark et al.; licensee BioMed Central Ltd. 2012
Published: 13 December 2012
MicroRNAs (miRNAs) have been known to play an important role in several biological processes in both animals and plants. Although several tools for miRNA and target identification are available, the number of tools tailored towards plants is limited, and those that are available have specific functionality, lack graphical user interfaces, and restrict the number of input sequences. Large-scale computational identifications of miRNAs and/or targets of several plants have been also reported. Their methods, however, are only described as flow diagrams, which require programming skills and the understanding of input and output of the connected programs to reproduce.
To overcome these limitations and programming complexities, we proposed C-mii as a ready-made software package for both plant miRNA and target identification. C-mii was designed and implemented based on established computational steps and criteria derived from previous literature with the following distinguishing features. First, software is easy to install with all-in-one programs and packaged databases. Second, it comes with graphical user interfaces (GUIs) for ease of use. Users can identify plant miRNAs and targets via step-by-step execution, explore the detailed results from each step, filter the results according to proposed constraints in plant miRNA and target biogenesis, and export sequences and structures of interest. Third, it supplies bird's eye views of the identification results with infographics and grouping information. Fourth, in terms of functionality, it extends the standard computational steps of miRNA target identification with miRNA-target folding and GO annotation. Fifth, it provides helper functions for the update of pre-installed databases and automatic recovery. Finally, it supports multi-project and multi-thread management.
C-mii constitutes the first complete software package with graphical user interfaces enabling computational identification of both plant miRNA genes and miRNA targets. With the provided functionalities, it can help accelerate the study of plant miRNAs and targets, especially for small and medium plant molecular labs without bioinformaticians. C-mii is freely available at http://www.biotec.or.th/isl/c-mii for both Windows and Ubuntu Linux platforms.
MicroRNAs (miRNAs) are a class of small, non-coding, single-stranded RNA molecules of 18-22 nucleotides. In various species, they play roles in gene regulation by targeting mRNAs at the post-transcriptional level . In plants, miRNAs are involved in organ development and environmental responses [2–4]. Although several miRNA and target prediction tools are available [5, 6], the number of tools customized for plant miRNA and target analysis is limited. Among them are microHARVESTER, a web server for identifying plant miRNAs ; the miRU , psRNATarget , and TAPIR  web servers for identifying plant miRNA targets; a web-based toolkit for the analysis of plant small RNAs ; and the miRTour web server for plant miRNA and target prediction . Most of these public web servers limit the number of input sequences and focus on only miRNA or target identification. Target-align  was recently proposed for plant miRNA target identification and developed as both web and command line versions. Even though several studies in computational identification of plant miRNAs and their targets are available [14–25], their methods were mainly presented as flow diagrams of connected programs (e.g., BLAST , UNAFold ). To follow the same steps, users need to install these programs, understand their usage, comprehend the meaning and format of the results, and have the programming experience for connecting them together.
Taking into account all these computational steps and criteria for plant miRNA and target identification [28–38], we developed C-mii, a standalone software package with graphical user interfaces for identifying, manipulating, and analyzing plant miRNAs and targets. C-mii is implemented as an all-in-one Java package weaving together sequence similarity search, secondary structure folding, automatic stem-loop identification and manipulation, and functional and gene ontology (GO) annotation. In addition, it comes pre-installed with databases of proteins, non-coding RNAs, and mature miRNAs. C-mii expects a set of nucleotide sequences (e.g., cDNAs, Expressed Sequence Tags (ESTs), Genome Survey Sequences (GSS)) in FASTA format as input. The identification steps are divided into miRNA and target identification pipelines. Users can customize parameter settings for each step of the identification, and filter and manipulate the results according to various biological criteria.
The target identification pipeline consists of four consecutive modules: sequence loading and validation, target scanning, miRNA-target folding, and target annotation. The sequence loading module is the same as that for miRNA identification, but accepts input sequences of longer lengths (≤ 20,000 nt). The target scanning module determines if input sequences contain the complementary sites to mature miRNAs of interest. Implementation of this module is similar to the homolog search except for the use of reverse-complement mature miRNAs as queries for BLAST. The miRNA-target folding module utilizes UNAFold to predict the secondary structure and free energy of miRNA:target duplexes. This module was introduced to C-mii to help refine the target identification result. The last module, target annotation, supplies function and gene ontology (GO) for the potential target sequences. Along the computational steps of both pipelines, users can customize parameters for their execution, select databases of interest, and explore and filter the results.
The pre-installed databases of C-mii are divided into three categories. First, the mature miRNA database, miRBase release 16, was incorporated for homolog search. Second, the non-coding RNA database, Rfam 10 with removed miRNAs, was introduced for removing other types of RNAs. Third, the protein databases, the UniProtKB/Swiss-Prot release 2010_12 and UniProtKB/TrEMBL release 2011_01, were incorporated for removing protein-coding sequences in the primary and precursor miRNA folding steps and for identifying gene functions during target annotation. These databases were pre-processed using the formatdb program in the BLAST package. Users can update these databases from our web site via a menu in C-mii. Furthermore, users can integrate their own protein and non-coding RNA databases into the system by following the pre-processing steps documented on the C-mii web site.
To ease its installation and usage, C-mii was designed as a complete package containing all required software, including BLAST, Java Development Kit (JDK), Perl, Python, UNAFold, and Ghostscript , which can be customized during installation. To explore conservation and co-evolution among the predicted and known precursors or mature miRNAs of an miRNA family, CLUSTAL W , MUSCLE  and Jalview  were pre-installed for performing and visualizing the multiple sequence alignment of selected identification results. In addition, we have deployed prefuse visualization toolkit , ICEpdf Viewer , and JFreeChart  for visualizing GO trees, the secondary structure folding of primary and precursor miRNAs, and the infographics, respectively.
C-mii is composed of two pipelines for plant miRNA and target identifications, which could be used autonomously or consecutively. The functionalities of these pipelines have been described with biological rationales and a running example (visit http://www.biotec.or.th/isl/c-mii/documentation.php under "C-mii running example section" to see screenshots of all steps).
Taking into account the computational steps and criteria for plant miRNA identification as described previously [28, 29, 31], the miRNA prediction menu consists of four consecutive submenus starting from sequence loading, homolog search, primary miRNA folding, and precursor miRNA folding. Users need to build a new project before uploading nucleotide sequences in a FASTA file. Sequences longer than 3,000 nt or containing characters other than A, T, C, G, U, and N will be excluded. As a running example, we built a TAIR10 cDNA project of 33,602 sequences. Due to sequence validation, 30,707 sequences remained as input for the homolog search.
An miRNA gene needs to be non-coding and have a stem-loop precursor in its secondary structure . The primary miRNA folding module helps users remove protein-coding sequences and other types of non-coding RNAs from input sequences, and predicts the secondary structure folding of primary miRNAs (pri-miRNAs) from the remaining sequences. The removal of unwanted sequence types is based on BLASTX and BLASTN against protein and non-coding RNA databases, selectable by users. Users can also adjust the E-value to limit the number of search results. The lower the E-value, the larger the number of sequences remains for secondary structure folding. C-mii sets the default E-value of BLASTN against Rfam as 1E-8 according to our previous benchmark . Users may also customize the folding temperature, maximum base pair distance, and maximum bulge or interior loop size of UNAFold.
Figure 2B shows the primary miRNA folding results with default parameter settings with the exception of the BLASTX E-value ≤ 1E-5. Sequences with a clickable BLASTN or BLASTX box are sequences that hit with other sequence types. By clicking on these boxes, users can explore their E-values and hit sequences. The "Only results" check box allows users to filter out these sequences. For the remaining sequences, users can interactively explore their secondary structures and minimal free energies (MFEs) predicted by UNAFold, minimal folding free energy indices (MFEI) , sequence lengths, and nucleotide and GC content. The "Mode" column provides two options for extracting stem-loop precursors from a secondary structure of a primary miRNA. C-mii decides the cleavage positions for users in Auto mode. The Manual mode allows users to specify the start and stop cleavage positions. From the results of our running example using the plus strand only, 567 Arabidopsis cDNAs in 124 miRNA families and 224 members remained. All these sequences were selected with Auto mode for the precursor miRNA folding module.
The precursor miRNA folding module helps users (1) extract the stem-loop structures from the secondary structures of pri-miRNAs, (2) remove stem-loop sequences that hit protein-coding sequences and other types of non-coding RNAs, (3) re-predict the secondary structure folding of the extracted stem-loop sequences, and (4) verify the predicted structures with previously suggested criteria. With the Auto mode setting from the previous step, the precursor miRNA folding module will cleave a pri-miRNA structure from the start position of the found homologous miRNA to the end position of its duplex miRNA* with two-nucleotide 3' overhangs. The extracted stem-loop sequences are screened against the protein and non-coding RNA databases again. UNAFold is then reapplied to the remaining sequences to predict the secondary structure folding of precursor miRNAs (pre-miRNAs). Users can customize the same set of parameters as in the primary miRNA folding step.
Figure 2C shows the precursor miRNA folding results with default parameter settings with the exception of the BLASTX E-value ≤ 1E-5. Based on previously reported criteria [28, 31], structures with multi-loops or mature miRNA sequences not located within one arm will be automatically removed by C-mii. Besides MFEIs, users can filter the results by restricting the number of two-nucleotide 3' overhangs, the number of mismatches between miRNA:miRNA* duplexes, the number of bulges, and bulge sizes as proposed in . Users can browse through the predicted secondary structures, select potential miRNAs based on filters, save them for target identification, and export them as an archive. Users may also perform multiple sequence alignments among the identified and known precursor or mature miRNAs of the same family to explore their conservation and evolution (Figure 2D). Using the plus strand only, 223 Arabidopsis cDNAs in 103 miRNA families and 197 members were finalized as potential miRNAs from 30,707 TAIR10 cDNAs (see System benchmarking and validation section for the detailed discussion).
C-mii's target prediction menu consists of four submenus: sequence loading, target scanning, miRNA-target folding, and target annotation. Users may build a new project for target identification only or continue the project from the miRNA identification pipeline. Users may also upload a new set of nucleotide sequences in FASTA format or reuse the uploaded sequences from the miRNA identification process. However, the acceptable length of an input sequence for target identification is extended to 20,000 nt. Besides sequence uploading, users also need to select mature miRNAs of interest from miRBase and/or from predicted mature miRNAs saved from the miRNA identification pipeline. In our running example, 434 sequences previously reported as miRNA-specific targets in Arabidopsis were uploaded to a new project and all mature Arabidopsis miRNAs from miRBase were selected for target scanning.
The miRNA-target folding module helps users refine their scanning results. It uses UNAFold to predict the secondary structures of miRNA:target duplexes, MFEs for hybridization, and binding positions of miRNA-target pairs. Users can specify the temperature that might affect miRNA:target duplex formation. Figure 3B shows the miRNA-target folding results of our running example using the default 37 ºC. From the results, users can determine potential miRNA targets based on the binding score from target scanning, the number of mismatches and G:U pairs between miRNA and its potential target from miRNA-target folding, mismatched positions, MFEs, and overlapped binding positions between the two steps. In our example, with the plus strand only, all 410 sequences remained for target annotation.
The target annotation module supplies function and gene ontology (GO) for potential targets selected from the previous step. Users can choose a protein database and customize the E-value and number of hits for BLASTX. Figure 3C shows the target annotation results of 400 out of 410 sequences using default parameter settings (see System benchmarking and validation section for the detailed discussion). From the results, users may explore GO annotation for a set of targets. By clicking on "Graph icon," C-mii calculates and visualizes the distribution of selected targets' GO IDs, colored by GO molecular function (F), biological process (P), and cellular component (C). From this view, users may also explore potential targets annotated with the same GO. The "Go View" allows users to investigate GO annotation of an individual potential target via a GO tree (Figure 3D). Users may also follow web links to public databases of known miRNAs and target functions.
Number of remaining input sequences from each step of miRNA identification on the four datasets
miRBase 16 (Arabidopsis only)
Rfam 10 (all plant RNAs except miRNAs)
miRNA identification steps
1. Sequence loading
2. Homolog search
3. Primary miRNA folding
4. Precursor miRNA folding
Number of TP, FP, FN, and TN of miRNA identification on the three datasets
Number of sequences
Number of identified miRNAs
miRBase 16.0 (Arabidopsis only)
Rfam 10 (all plant RNAs except miRNAs)
Number of remaining input sequences from each step of target identification on the previously reported miRNA-associated target sequences of Arabidopsi s
Step/Number of remaining sequences
Number of input sequences
1. Sequence loading
2. Target scanning*
3. miRNA-target folding
4. Target annotation**
Time usage for each step of miRNA and target identifications on TAIR10 cDNA dataset with varied number of threads running
miRNA identification steps
Primary miRNA folding
Precursor miRNA folding
Target identification steps
This paper presents C-mii, a standalone software package for computational identification of plant miRNAs and targets. C-mii has been implemented as all-in-one Java package with following distinguishing features. First, it comes with graphical user interfaces of well-defined pipelines for both miRNA and target identifications, with reliable results. Second, it provides a set of filters allowing users to reduce the number of results corresponding to the recently proposed constraints in plant miRNA and target biogenesis. Third, it extends the standard computational steps of miRNA target identification with an miRNA-target folding module and GO annotation. Fourth, it supplies bird's eye views of the identification results with infographics and grouping information. Fifth, it provides helper functions for database update and auto-recovery to ease system usage and maintenance. Finally, it supports multi-project and multi-thread management to improve computational speed. With these features, C-mii is a very useful software package that can help accelerate the study of plant miRNAs and targets by plant biologists.
Project name: C-mii: A tool for plant miRNA and target identification
Project home page: http://www.biotec.or.th/isl/c-mii
Operating system(s): Windows and Ubuntu Linux 9.10 or higher
Programming language: Java and Python
Other requirements: -
License: GNU GPL
Any restrictions to use by non-academics: license needed
The authors would like to thank Drs. Anan Jongkaewwattana, Samaporn Teeravechyan, and Sithichoke Tangphatsornruang for editing the manuscript and giving helpful comments on the results. This work was supported by grant FC0033 B21 (SPA B1-1) from Cluster Program Management Office (CPMO), National Science and Technology Development Agency (NSTDA), Thailand.
This article has been published as part of BMC Genomics Volume 13 Supplement 7, 2012: Eleventh International Conference on Bioinformatics (InCoB2012): Computational Biology. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcgenomics/supplements/13/S7.
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.