Closed bhofmei closed 5 years ago
Hi, You can ignore the runtime warning. It occurs when division by zero is encountered. NA means that HOME did not find any significant DMR for that sample comparison. Will update that on the documentation. Also, if you are getting too many NAs it might be worth trying running HOME pairwise for multiple samples. So, if you run the same sample file path used for timeseries in normal HOME pairwise mode you will get DMRs for each comparison and then you can merge them to get the final DMRs for all sample comparison. Please let me know if you have any other issues/doubts. Thanks, Akanksha
Thank you for the explanation. I was guessing that NA meant it wasn’t signficant but then I didn’t understand why the DMR would be included in the output when it was NA at all comparisons.
I had tried the pairwise technique before but was struggling to elegantly merge the DMRs afterwards and comparisons afterwards. It appears that bedtools merge would work to find a master list of DMRs and maybe say what comparison it came from, but it’s still not elegant since the new DMR boundaries, and potentially multiple testing correction, could change the statistical significance. My data is definitely time series, so if I were to use this technique, would you suggest merging from all comparisons or only the timeseries-relevant ones?
On Oct 11, 2018, at 7:19 PM, Akanksha2511 notifications@github.com wrote:
Hi, You can ignore the runtime warning. It occurs when division by zero is encountered. NA means that HOME did not find any significant DMR for that sample comparison. Will update that on the documentation. Also, if you are getting too many NAs it might be worth trying running HOME pairwise for multiple samples. So, if you run the same sample file path used for timeseries in normal HOME pairwise mode you will get DMRs for each comparison and then you can merge them to get the final DMRs for all sample comparison. Please let me know if you have any other issues/doubts. Thanks, Akanksha
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ListerLab/HOME/issues/27#issuecomment-429153542, or mute the thread https://github.com/notifications/unsubscribe-auth/APWCOjyMm1MnW3KRU7hZOVZlVZQnQIATks5uj9IdgaJpZM4XX41j.
Hard to say, without looking into the data. I would suggest merging for all comparisons, otherwise, how would you find the time series-relevant ones?.
Okay thank you! I'll try that. FWIW, it would be awesome if you added some sort of try-catch or error check for the the divisible by zero error. Otherwise it's difficult to tell if the program actually finished and finished correctly.
yes sure. Thanks
I keep getting this error running timeseries with CHH context. Just wondering if you know what is causing it or how to fix it.
Also, can you update the documentation/README to say what
NA
means in the timeseries combn-1 columns? Current documentation says these cells havestart:end:state:delta
but most of mine areNA
even with high confidence scores. Just trying to understand.