Classifier

In this excercise, a lot of concepts will come together to form a basic classifier.

Problem Description

Build a CLI tool classifier for sequence classification of reads against a database. The output should be a .tsv file, containing the classification in sorted order. E.g., the database sequence with the best classification should be first in the list.

Expected command:

classifier --fastq <reads.fastq.gz> --db <db.fasta>.

Expected example output:

sequence_x  1.0
sequence y  0.95
sequence z  0.94
...

Where sequence_x was fully covered, sequence_y was 95% covered, etc. Define the score as 1.0 if 100% of the FracMinHash kmers from a sequence is contained in the reads.

Checklist

• Implement FracMinHash generation for a given byte slice &[u8].
• Generate FracMinHashes for all reads.
• Calculate the score for each db sequence.

Suggested Rust Crates

• Clap - argument parsing.
• Needletail - reading fasta files.

Code Examples

• Take inspiration from code examples in this book.

Extra Credits

• Only allow input files with extensions .fasta, .fa, .fsa, .fna for the database file and .fastq, .fastq.gz for the reads.
• Graceful error handling.
• Replace Needletail with Bio + Rayon to enable multi-threading.
• Add a coverage metric to show the mean FracMinHash coverage for each sequence. Hint - don't use a HashSet, but rather a HashMap to count the FracMinHash read kmers.