In the [Prediciton page], you can start a new prediciton by following steps:
Step 1, Upload your methylation peak/site in BED6 file format.
Step 2, Select a cell type model that is most similar to the cell type of your interest.
Step 3, Click the submit button to run the prediction task.
Step 4a, Bookmark the task query result page, and check the progress later.
Or you can save the task ID, and query your previous task at [Here].
Reading-m6A accepts a six-column BED file (BED6), where the columns read chromosome, methylation site start (0-based), methylation site end, methylation site name, methylation site socre, and strand, respectively.
You can find a sample BED6 file at [Here].
Typically half an hour to an hour. While the prediction model does not take that long, intensive computation is required to prepare the genomic features of each methylation site, which is relatively time consuming.
It is very likely that your input does not comply with the BED6 format, please see the sample file at [Here]
Please also note that the BED6 file is 0-based, that is, for example, the start and end point of even a single nucleotide site at Chr11:123456 should read 123455 and 123456, respectively.
Yes, the number of readout types that could be covered by Reading-m6A depends on the readouts covered by high-throughput experiments. In our work, we have profiled mRNA half-life, translation efficiency and alternative splicing for five representative cell types (A549, HEK293T, hESC, HUVEC and JURKAT), so there will be five readouts to predict. However, for other cell types, only public data derived translational efficiency profiles are available, so only translational efficiency upregulation and translational efficiency down-regulation will be predicted for these cases.
All readouts follow m6A-normal vs. m6A-disrupted comparisons. For example, a translationally up-regulated gene would have a higher translation efficiency in normal cells than in m6A(METTL3) knockdown cells.
Our observations in high-throughput profiles suggest that the overall trend of half-life readout is toward down-regulation, unlike translation and exon inclusion, which are more balanced. Therefore, we do not consider half-life up-regulated genes as there is not enough sample to train our models.
No. Our model is a classification model, so the prediction score is about the probability of a readout, not about the fold change. You may want to combine the prediction with other evidence such as differential methylation, differential gene expression, etc. to narrow down your candidate gene list.