FAQ

FAQ

My sample is missing from the results. The sample name is everything before the first underscore of the file name. Two files whose names start the same way collapse into one sample (the extra long-read file is reported and ignored); a long-read file named unclassified* is dropped on purpose (unbinned barcode reads). In hybrid mode, short-read files whose name matches no long-read sample are reported and skipped.

Why did short-read polishing not run? It needs a pair (_R1 + _R2) and an assembly for that sample. Single-end short reads polish nothing (Polypolish and pyPOLCA want pairs); they are only assembled in short-read-only mode. Also check --short-polisher was not none.

Can I assemble PacBio reads? The long-read pipeline is tuned for Nanopore (Flye --nano-hq, Medaka). PacBio HiFi will assemble, but use --long-polisher none (Medaka models are Nanopore-only) — or better, a HiFi-specific assembler.

Medaka crashes or hangs. Medaka runs one sample at a time by design (parallel instances crash). If a single run still fails, the usual culprit is a broken samtools/bcftools in the environment: run samtools version, and recreate the environment if it errors. Also check the --model matches your basecaller — hashbrown --list-models.

How do I rerun just one step? Delete that step's done_* flag and its output folder, then rerun the same command — see Usage § Resuming.

It says a program is not installed, but I don't use that step. HashBrown only checks the programs your flags need. If it names one, some requested step uses it — e.g. nucmer/mummerplot/gnuplot belong to the long-read QC, bwa/polypolish/pypolca to hybrid polishing.

Out of memory. Lower -p (samples in parallel) before lowering -m: each parallel sample multiplies the footprint. SPAdes splits -m across the parallel samples; SKESA takes it per sample.

Where is the final assembly? The deepest folder that exists for the sample: 6_polished_short/, else 4_polished_long/, else 3_assembled_long/ (short-read-only runs: 2_assembled_short/). See Outputs.

How is this different from BGA / BGA_short? One tool, one environment, pluggable polishers, plus the fixes listed in the changelog. The old repositories stay up but are superseded.