Using eventalign to align raw RNA signal with high error

[awjga]

Hi Jared,

I am using eventalign to align signals from direct RNA sequencing. However, as the RNA that I am sequencing is expected to have a signal that is different from the current in the kmer models, I am hoping that I can relax the parameters for nanopolish, so that more reads can be aligned.

I have looked at under the part for RNA: nanopolish_squiggle_read.cpp

 // Filter poor quality reads that have too many "stays"
    if(!this->events[strand_idx].empty() && this->events_per_base[strand_idx] > 5.0) {
        g_qc_fail_reads += 1;
        events[0].clear();
        events[1].clear();

And I am thinking of replacing 5.0 to 10.0

 // Filter poor quality reads that have too many "stays"
    if(!this->events[strand_idx].empty() && this->events_per_base[strand_idx] > 10.0) {
        g_qc_fail_reads += 1;
        events[0].clear();
        events[1].clear();

Will there be any adverse effect for the alignment of the signal? Or is it is not advisable to do this?

Thanks! Ash

[jts]

Hi,

Sorry for the slow reply, I was traveling last week.

I don't think it is safe to make this change. This checks that the segmentation of the read is roughly as expected. Unless the change you made drastically changes the speed the RNA goes through the pore, you probably don't want to use the reads that fail this condition in your analysis.

Jared

[awjga]

Hi Jared,

Thanks for the reply. Then is it advisable to change MIN_CALIBRATION_VAR to a bigger value?

And I have noticed something from the raw signal.

Do you see this type of signal? There seems to be some weird spikes that is way above and below the expected range of the current?

I am thinking that if these outliers are removed, the accuracy of the data can be improved?

Thanks! Ash