"Error: length(barcode.i) == 1 is not true" and execution halted

[bensworth]

While waiting for the new deployment, we started using the virtual machine. The demos worked fine, but with our data, we get the error:

Error: length(barcode.i) == 1 is not true

followed by “Execution halted.” However, the executable isn’t killed. It’s not clear if it is still running, but it does not appear to be doing anything.

We start with one data file, and the the error comes up ~30 minutes after the code starts running. There is no other readout before this comes up. It happens after all of the "trimmed" and "corrected" files are generated in the Data folder, but the demultiplexedReps folder is still empty. Unlike in the demo run, the corrected reads and trimmed reads do not match in the Data folder. Our barcodes are only different by an edit distance of 3, and one suspicion is this could be part of the problem.

It is likely this is not an “issue,” but any insight into what this error means/how to address it would be greatly appreciated!

[cnobles]

This error message is meant to catch non-unique barcodes. If you've used the same barcode on different samples, then this statement will be FALSE, even with different linkers. While the pipeline technically does do dual-barcode demultiplexing, it still requires that the "barcode" or index read sequence be unique for each sample. This is certainly something we will be changing in an updated version.

As a fix in the short term, try splitting your sampleInfo file into groups that only have unique barcode sequences. For instance, if you had 16 samples and recycled 8 barcodes, you could split the sampleInfo file into two files, each with the 8 unique barcodes.

The output data can then be combined in the database (sql or sqlite) or combined in an R-session.

It's likely that the error won't show up for some time as the program first tries to error-correct the barcode sequences then demultiplex with them.

[cnobles]

Also, you should try out the INSPIIRED-deployment branch. John had worked up a new environment and updates to libraries. Let us know how that turns out.

[bensworth]

After assigning unique barcodes to each sample we were able to get past the previous error, but are encountering a new one later in intSiteCaller:

Can't open NA to read: No such file or directory
[1] TRUE
Error in strsplit(alignFile, "/")[[1]] : subscript out of bounds
Calls: alignSeqs
Execution halted

The last two lines in intSiteCaller.log are,

     Waiting for alignSeqs-1.done ...
     Waiting for alignSeqs-0.done ...

The run directory contains alignSeqs-1.done but not alignSeqs-0.done.

Related to the barcodes, simply using unique barcodes did not bypass the error barcode(length.i) == 1 is not true. Only after manually replacing our barcodes with the exact barcodes used in the demo set was that error avoided.

UPDATE: These are screenshots of the same step in the demo run and my sample run. My sample run counts barcoded reads and does nothing more before moving to the next sample. The demo run counts barcoded, LTRed, linkered reads, etc. and does some other things before moving to the next sample. See attached screenshots for the demo and my sample terminal printout, respectively, and the .log file (identical in each sample folder).

intSiteCaller.log

[helixscript]

All of INSPIIRED intermediate files are 1 based.
Log entries and intermediate files that reference from 0, such as 'alignSeqs-0.done' is indicative of an error that was not caught by the software where it should of halted. Such issues can arise if all reads are removed by the software filter -- ie. all reads have match to the vector and were removed. We will add additional monitoring.

[helixscript]

I updated the deployment branch.

https://github.com/BushmanLab/INSPIIRED/tree/INSPIIRED-deployment

I locked down some more library versions so that Conda does not try to
auto-update additional libraries which subsequently breaks the code base.

I created a new INSPIIREDv2 environment and changed the install procedure
which is documented in the Github readme.

Please first remove the old INSPIIRED environment before attempting to install the new version.

ie.

%> cd INSPIIRED
%> export INSPIIRED=$(pwd)
%> ~/miniconda3/bin/conda remove --name INSPIIREDv2 --all
%> ~/miniconda3/bin/conda update conda
%> ~/miniconda3/bin/conda env create -f bin/INSPIIRED.conda.yml
%> source ~/miniconda3/bin/activate INSPIIREDv2
%> Rscript bin/setupR.R
%> Rscript bin/setupINSPIIRED.R

I've tested this up to the point of uploading the identified sites to the provided sqlLite db.
We have new report making software which I will release shortly.

Are you working with a CPU cluster (bsub/qsub) or are you running the software in a single CPU?