Methodology

Under the hood, AuriClass conceptually works as follows:

flowchart TD
  A[Argument parsing & input validation] -->|FastA data| B_fa[Create FastaAuriclass object];
  A --> |FastQ data| B_fq[Create FastqAuriclass object];
  B_fa --> C_fa[Sketch input FastA];
  B_fq --> C_fq[Sketch input FastQ];
  C_fa --> D[Run mash dist & select closest sample];
  C_fq --> D;
  D --> genome_size["Check whether genome size
                    is within expected range"];
  D --> E[Check if sample is close enough to any reference sample];
  E --> |Too distant| F_fail["FAIL: not Candida auris
                                Further QC skipped"];
  E --> |Close enough| F[Check is sample is closest to C. auris];
  F --> |Too distant| G_fail[WARN: other Candida/CUG-Ser1 clade sp.];
  F --> |Close enough| G[Select closest clade];
  G --> H[Final quality checks];
  G --> output
  H --> output

  genome_size --> output[Output]
  F_fail --> output
  G_fail --> output

Note

Whether input files are FastQ or FastA is guessed by default based on pyfastx parsing. This can be overridden by specifying --fastq or --fasta.

Steps that differ between FastQ and FastA data:

• Sketching is performed differently: filtering for minimal kmer coverage is required for FastQ.
• Genome size is estimated for FastQ using mash sketch, and parsed from FastA using pyfastx

FastqAuriclass and FastaAuriclass classes are based on BasicAuriclass which handles most of the attribute setting. The specific classes define some format-specific methods and how all methods should be called by the run() method.