threshold and motifs

[TiBi-MU]

Hello,
In the modkit summary, the threshold is calculated as a 10% percentile. It's not clear to me of what exactly. Can you please clarify that for me? Next, after visualizing the bedgraph in IGV, for example, when 6mA is detected, some positions of C or G are also marked, is this ok? Finally, is there any way to extract methylated motifs from the data? Thanks!

[marcus1487]

The 10th percentile is taken over modified base call probabilities. Note that this is the probability of the called base whether it be canonical or a modified base. Some options (mapped only, bed regions, sampling etc) can effect exactly the set of modified base calls included in the percentile computation. The reason for the 10th percentile instead of a fixed probability is that different models and samples (primarily driven by modified base content/density) can show different probability distributions. If you are running many experiments with similar samples it would be recommended to determine an ideal threshold (using modkit sample-probs) and apply the constant threshold over all samples.

6mA calls at non-A reference positions are due to either SNPs in the sample or basecalling errors. These can be filtered out with the --motif option to pileup or via a bedtools intersect after running pileup.

We do not have a tool released for motif detection at the moment, but have an internal prototype and hope to have this released in the coming months.