Ok, the fastq data formatting is the problem.
Compare the successful runs (with correct fastq formatting) and the two unsuccessful runs (with incorrect formatting for the forward reads, each).
Those forward reads should be reprocessed in the same way that the reverse reads were. It looks like the source data is slightly different between the forward/reverse reads for the failed jobs. The reverse reads were sourced/prepped the same as both the forward/reverse reads in the successful Bowtie2 job.
The data prep/trimming is being done outside of Galaxy, before Upload. SRA reads can also be extracted directly and QA steps done all within Galaxy as an alternative.
Tool choices to get SRA reads:
- NCBI SRA Tools > Download and Extract Reads in FASTA/Q format from NCBI SRA
- Get Data > EBI SRA (use the “Galaxy” fastq links)
Basic read QA in Galaxy is covered in the first tutorial, with more in the many of the analysis-specific tutorials from the Galaxy Training Network (GTN):
- NGS logistics - this is an introduction to Galaxy’s functionality for the analysis of Next Generation Sequencing data.
- GTN Galaxy tutorials
Reference Post: Troubleshooting resources for errors or unexpected results