Why doesn't pbtranscript.py classify call reads of inserts for films with 2 to 5 rds

[cklopp]

I've processed 8 pacbio Cells corresponding to 3 different IsoSeq libraries with the pbtranscript.py pipeline.
In the classify step of the procedure about 20% of the films (ZMW) do not produce a read of insert (RoI).
When I check the number of reads per film for reads giving and not giving RoIs I get the following result.



Which shows that films with 2 to 5 reads do not produce RoI or produce much less RoIs than other films.

Any idea why?

[Magdoll]

I would like some clarification on what you mean by "not producing a RoI".

The Iso-Seq classify steps are:

--- using the CCS algorithm (which is generic and used for many things in addition by Iso-Seq) to generate RoI reads (in the future, they may be called CCS reads again, sorry for all the naming changes!)

--- look at the RoI reads to identify 5' and 3' cDNA primers on the ends. It then "classifies" those RoI reads into full-length (has both 5' and 3' primer and polyA tail), and non-full-length (missing at least one of the criteria).


When you say "no RoI", do you mean:
(a) there was no RoI/CCS read for that ZMW.
or
(b) it was not full-length

Also, are all the libraries the same size? What is the avg. transcript length in these libraries?

I'm not entirely sure how I would explain what you observe (since I've not seen this myself). I did a # of passes vs RoI full-length detection survey a while back and it's different from what you see and is closer to what I'd expect:

https://github.com/PacificBiosciences/cDNA_primer/wiki/Comparison-of-Reads-Of-Insert-parameters-and-full-length-detection

Also for reference, here is a tutorial on using classify. It explains the parameters in detail:

https://github.com/PacificBiosciences/cDNA_primer/wiki/RS_IsoSeq-%28v2.3%29-Tutorial-%231.-Getting-full-length-reads

And another wiki to explain what to expect from classify output:

https://github.com/PacificBiosciences/cDNA_primer/wiki/RS_IsoSeq-%28v2.3%29-Interpreting-Classify-and-Cluster-Output

[cklopp]

"no RoI" means (a) there was no RoI/CCS read for that ZMW.

I've simply compared the ZMW names in the initial subreads file with the names in the RoI file.

The libraries are of three sizes (1-2kb, 2-3kb, 3-6kb). The average lengths are respectively 2kb, 2.5kb and 3.2kb.

[ndelaney]

Your reads are likely being filtered out by one of the criteria used (and which can be set as options to the command).

If using CCS2, you should see a report such as ccs_report.csv that gives a break down of what reads were filtered and why. If using a more recent version of CCS1, after the program finishes running it will print a report that indicates the yield loss due to various filters. If you can report either of these results here I can give more guidance.