CentroidFold Web Application Help

Back to CentroidFold

1. Why Use CentroidFold?

The major advantage of this server is that it employs our original CentroidFold software as its prediction engine which scores the best accuracy in our benchmark results. The details of the CentroidFold software and its performance evaluation are described in our recently published paper.
The following figures show the results of our performance comparison among CentroidFold and other popular programs.

The above figure shows our performance comparison results among CentroidFold version 0.05, RNAfold version 1.7, CONTRAfold version 2.02, SimFold version 1.10 and Sfold version 2.1. The comparison is carried out for the S-151 RFAM dataset used in Do et al., 2006 which contains 151 RNA sequences, each of which was taken from a different family in the RFAM database. The green curve representing CentroidFold out performs the other contenders.


The above figure shows another performance comparison among CentroidFold version 0.05, RNAalifold version 1.7 (*1) and McCaskill-MEA. The comparison was carried out for the dataset used in Kiryu et al., 2007, which contains 85 reference alignments of 10 sequences taken from 17 RNA families in the RFAM database. Again, CentroidFold (green) scores the best performance among other contenders.
(*1 A new version of RNAalifold has recently been released. See their web site.)

2. Interface

This is a screenshot of the CentroidFold top page.

You can find a link to this page shown as "Help page".
Another link shown as "more details can be found here" leads you to several other pages including software download.
Several instant examples follow the text saying "input sequences: like ...".
There are three single sequence examples:

1. Hummerhead ribozyme
2. tRNA
3. H/ACA snoRNA

The above figure shows a screenshot with a sample input sequence placed in the text area.

There are four multiple-alignment examples:

1. traj 5'UTR
2. Qrr RNA
3. Histidine operon leader
4. U11 spliceosomal RNA

The above figure shows a screenshot with a sample multiple-alignment placed in the text area.

Please click on these labels to try CentroidFold with a sample input then click on "execute CentroidFold" button to perform prediction.


The above figure shows a screenshot with a sample result for Hammerhead ribozyme.
The results are shown in several ways: the input sequence with base-pair notations, a popular secondary structure graph, and its PDF version can be downloaded (click "Display the PDF file").
Each predicted base-pair is colored with the heat color gradation from blue to red corresponding the base-pairing probability from 0 to 1, where the base-pairing probability is the probability that a pair of bases forms a base-pair via hydrogen bonds in their secondary structures, and can be interpreted as confidence measure of predicted base-pairs.


The above figure shows a screenshot with sample result for traj 5'UTR.

The results are shown in several ways: the input multiple-alignment with base-pair notations, a popular  secondary structure graph, and its PDF version can be downloaded (click "Display the PDF file").

You can select advanced options as the follows:

We provide the CONTRAfold model and the McCaskill model as the choice of the model for calculating base-pairing probabilities. The gamma is a weight parameter for predicted base-pairs for controling the trade-off between specificity and sensitivity of base-pairs. The larger gamma you use, more base-pairs are predicted. More details are described in our original paper.

3. Input Format

CentroidFold accepts the following sequence formats:

• Sequence-in-a-line: One sequence in a line. Please do not fold a long sequence by a "carriage return" which results in several shorter sequences.
• FASTA: The famous header-sequence combined format.
• CLUSTALW: The major multiple-alignment file. When this format is found in the input, CentroidFold predicts a common secondary structure.

Please paste your query sequence in one of the listed formats.

4. Misc.

The sequence length is limited up to 2000 bases.
The following figure shows CentroidFold computational time per sequence length.


Back to CentroidFold