Genome recovery

Genome recovery groups assembled contigs into candidate genomes, refines the candidate sets and evaluates the resulting MAGs. Reads are still important: mapping them back to the assembly provides coverage information used by binning and abundance calculations.

Basic usage

aviary recover \
  --assembly assembly/final_contigs.fasta \
  -1 sample_R1.fastq.gz \
  -2 sample_R2.fastq.gz \
  --output recovery_run \
  --max-threads 16 \
  --n-cores 32

How the stage behaves

Aviary prepares coverage inputs, invokes the enabled binners, refines candidate bins and combines evidence into a final set. Quality assessment and taxonomy then add evidence needed to interpret each MAG. A binning algorithm producing a FASTA file does not by itself establish that the genome is complete, uncontaminated or correctly classified.

Important controls

--min-bin-size excludes candidate genomes smaller than the configured number of bases from later processing. Binner selection options change which upstream algorithms contribute candidates. GPU variants require a compatible GPU and the corresponding Aviary environments; enabling a GPU flag does not make every stage GPU accelerated.

SemiBin2 multi-sample mode accepts multiple assemblies and should be used only with inputs organised for that mode. See the exact validation and accepted values in the recover reference.

Interpreting results

Use bins/final_bins/ as the recovered FASTA collection and bins/bin_info.tsv as its main summary. Retain the run's database versions and software environment because taxonomy and quality estimates can change with reference data and dependency versions.

Continue with annotation, the output reference, or reproducibility guidance.