Fiter for a minimum mean read quality did not work when "cut front" and "cut tail" were turned on

[wanyuac]

Software: fastplong v0.3.0

Command:

fastplong -i isolate1_raw.fastq.gz -o isolate1_processed.fastq.gz --cut_front 15 --cut_tail 15 --cut_window_size 4 --mean_qual 15 \
--qualified_quality_phred 15 --unqualified_percent_limit 50 --length_required 1000 \
--json isolate1_fastplong.json --html isolate1_fastplong.html --report_title "Fastplong report: isolate 1" --thread 4

Nanoplot visualisation of the quality of output reads:

It seems that the filter for the mean quality of Q15 did not work. I would much appreciate it if anyone would kindly share any insights on this issue.

[lsterck]

Not 100% but I think the discrepancy might be due to the fact that fastplong uses that mean in the specified window (so not the whole read) and that nanoplot plots you the avg quality over the whole read.
If the window approach of fastplong encounters a window where it meets the mean_qual threshold it will stop.

by using the mean_qual quality score option you likely get the desired result (and visible on the nanoplot plot) == filtering on av read quality

[ndombrowski]

I wanted to ask if there has been any update on how --mean_qual is calculated. I encounter the same discrepancy: when filtering reads with -l 1300 --length_limit 1600 --mean_qual 15. As you see below, the resulting reads, when visualized with NanoPlot, still include many reads with low average quality:

This seems inconsistent with the Fastplong HTML output, which reports much higher raw quality for the same dataset (and also mean values that are too high for what I would expect for this kind of data):

I suspect the discrepancy arises from how the average quality is calculated. NanoPlot appears to compute the average by converting PHRED scores to error probabilities, averaging those, and converting back to a PHRED score (code, discussion).

Could you clarify whether --mean_qual in Fastplong is calculated as a arithmetic mean of PHRED scores, or in a way comparable to NanoPlot?