Welcome to SRAMP prediction server

N6-methyladenosine (m6A) is a post-transcriptional methylation modification that widely presents at the adenosine bases of RNA transcripts. This modification has been suggested to be involved in the regulation of the degradation, subcellular localization, splicing and local conformation changes of the RNA transcripts. In spite of the functional importance of the m6A, its catalyzing mechanism is still not fully understood. For example, previous experiments have proven that m6A preferably occurs on the DRACH motif (i.e. [AGU][AG]AC[ACU] in regular expression style). Nevertheless, high-thought sequencing data have demonstrated that in the mammalian transcriptome, only a small fraction of such motifs are really modified, suggesting additional sequence and structural features which determines the m6A modification sites.

Thanks to the latest development of high-throughput techniques that unambiguously identifies mammalian m6A sites in single-nucleotide resolution, now we are able to extract and integrate the sequence and predicted structural features around m6A sites under a machine learning framework and build a mammalian m6A sites predictor: SRAMP (sequence-based RNA adenosine methylation site predictor). SRAMP achieves promising performance both in cross-validation tests on its training dataset, and in the rigorous independent tests. Another highlighting trait of this predictor is that only RNA sequences are required when running a prediction and no external -omics data are loaded. Therefore, SRAMP would serve as a useful tool to predict m6A modification sites on the RNA sequences of interests.

June, 2016, a simplified downloadable version of SRAMP was made available;
December, 2015, prediction models were updated, tissue-specific predictions were enabled;
Octorber, 2015, RNA secondary structure visualization was enhanced;
September, 2015, mature mRNA-based prediction mode was established;
July, 2015, full transcript-based prediction mode was established;

Contact us
Dr. Yuan Zhou
38 Xueyuan Rd, Department of Biomedical Informatics, Peking University Health Science Center, Beijing 100191, China
Email: soontide6825@163.com

Zhou Y, Zeng P, Li YH, Zhang Z & Cui Q (2016) SRAMP: prediction of mammalian N6-methyladenosine (m6A) sites based on sequence-derived features. Nucleic Acids Research, 44(10):e91.

Please choose one preferred prediction model:

Full transcript mode

(Click here for an example sequence)

No error message

Mature mRNA mode

(Click here for an example sequence)

No error message

1. What is required for SRAMP run predictions?

SRAMP requires nucleotide sequence only for running a prediction. There are two prediction modes available, each accepts different kinds of sequences, which are specified below:

2. Should I analyze RNA secondary structure?

Through analysis of RNA secondary structures, the text representation and graphical visualization of the local RNA secondary structures around the predicted m6A sites will be provided. On the other hand, it takes MUCH MORE TIME to run a prediction while analyzing RNA secondary structure. For a quick prediction, the 'analyzing RNA secondary structure' option is by default disabled. But this option can be easily enabled during the submission of a new prediction task.

3. How can I interpret the 'Structural Context' string in the result page?

The structural context string represents the secondary structure status at each position flanking the predicted m6A site in text format. Its alphabet reads: P- Paired, M- Multiple loop, I- Interior loop, B- Bulged loop, H- Hairpin loop.

4. How are the very high/high/moderate/low confidence m6A sites decided?

The thresholds for very high/high/moderate/low confidence m6A sites correspond to the threholds achieved 99%/95%/90%/85% specificities (in other words, had 5%/10%/15% false positive rate) on cross-validation tests, respectively. We suggest users who wish to ensure low false positve rate that consider the predicted m6A sites with at least high confidence.

5. Should I wait until the prediction job is finished?

Not necessary. You would like to bookmark the processing page and look after the results later. Nevertheless, please note that all of the result from a submission will be automatically DELETED AFTER 72 HOURS.

6. What prediction model should I choose? Generic or tissue-specific?

The generic prediction model is competence for usual prediction tasks. If it doesn't produce satisfactory results, you may wish to try tissue-specific prediction models.

7. How can I predict on multiple sequences?

Due to the limitation of our computation resources, we require one sequence per prediction on our webserver. Nevertheless, A simplified, downloadable version of SRAMP is available on the 'Download' webpage of this server.