Introduction


N6-methyladenosine (m6A) is one of the most common RNA post-transcriptional chemical modifications. It was found to have influence over diverse biological regulations such as RNA stability, heat shock response, and circadian clock etc. Current researches focus more on the overall functions or regulations that m6A involves in, the separated function of each methylation site is still ambiguous. Although regulation roles of several specific methylation sites have been detected, it is cost expensive to test the site functions with lab experiments. Instead, if functions of each methylation site can be predicted by the statistical evidence such as strong correlation between site pairs, the coverage of function prediction will be much higher.


1446 human m6A sites were annotated by Gene Ontology (GO) BP terms with a co-methylation network. We suppose that if both methylation sites are hyper- or hypo-methylated together across various samples, this methylation site pair will be considered as the co-methylation pair, and this site pair will also be of biological interest. After the raw data processing (reduce the systematic biaes) and site filtering step, the network is constructed based on 40 samples from 10 public human MeRIP-Seq data sets, with the guilt-by-association principle by two different methods: the hub-based method and the module-based method.


In the hub-based method, functions of methylation sites are determined by the enrichment result of their neighbor sites in the network, while in the module-based method, the site functions are enriched with the modules clustered by the MCL algorithm. According to our calculation, the performance of the functional prediction of sites with hub-based method in the real network is higher (about 15%) than the random network, which proves that the prediction in the real network should be of biological significance. In addition, the overlap terms of both predicted methods in the real network (27) is much higher than the random network (3.2), which indicates that the predicted functions annotated by the module-based method are credible in some level.


The annotation result can be viewed in this m6Acomet. You can:

1) Quick search the name of gene or function using a search query on the Home page.
2) Visualize the table which contains the overall summary of predicted human m6A sites on theTable page, you can click on the site name or the term count in the table to view the detailed information of each site.
3) Download this table on the Download page.