Is your feature request related to a problem? Please describe.
When working with low-coverage and rare samples, a user will rerun a barcode in the same flowcell to increase the number of reads. Fastq files will then be merged into the same barcode directory and run through the pipeline. It is challenging to know if barcodes have enough reads to complete assemblies. Sometimes low coverage samples will cause Unicycler to throw an error re: miniasm.
Describe the solution you'd like
We should be doing some basic evaluation of read depth for barcodes at the input check stage of the workflow, these could throw a warning for samples that may not pass a specific threshold (don't know what this would be). Additionally, we could implement a no-miniasm implementation of unicycler for samples that may not meet this threshold to give it the best chance of success. Things to consider:
What would be a reasonable threshold for read depth?
What other qc checks should we do, as read depth is not the only important metric for genome assembly?
What impact does skipping miniasm have on the assembly and does this matter in our application?
How can we implement this in a way that is relatively straight forward and easy to understand for an end user?
Describe alternatives you've considered
None really, except skipping and recalling samples with low coverage.
Is your feature request related to a problem? Please describe. When working with low-coverage and rare samples, a user will rerun a barcode in the same flowcell to increase the number of reads. Fastq files will then be merged into the same barcode directory and run through the pipeline. It is challenging to know if barcodes have enough reads to complete assemblies. Sometimes low coverage samples will cause Unicycler to throw an error re: miniasm.
Describe the solution you'd like We should be doing some basic evaluation of read depth for barcodes at the input check stage of the workflow, these could throw a warning for samples that may not pass a specific threshold (don't know what this would be). Additionally, we could implement a no-miniasm implementation of unicycler for samples that may not meet this threshold to give it the best chance of success. Things to consider:
Describe alternatives you've considered None really, except skipping and recalling samples with low coverage.