Detecting SARS-CoV-2 variants from the mixed virus population

Question

I have a fastq file from Nanopore sequencing data that contains reads from both the UK and South Africa variants of SARS-CoV-2.

The variants are identified by three key mutations in the receptor-binding domain of Spike.

Aim Firstly, I want to determine the number of reads then belongs to UK and South Africa variant. Then, I will create a bam file from the fastq file using the SARS-CoV-2 reference genome.

Questions

1. How do I determine the reads belonging to each variants?
2. Are there any tools available to classify the reads based on the pre-defined mutation criteria?

Answer 1

Based on a quick request from someone else, I wrote some code a couple of years ago for assigning read groups based on variants at particular locations; this was set up for HIV variant splitting, so I expect it should also work on SARS-CoV-2 nanopore reads as well:

https://gitlab.com/gringer/bioinfscripts/-/blob/master/samVarSplitter.pl

It processes a SAM stream [i.e. in text mode] using Perl, so might be a bit slow, but should be fine for a few hundred thousand sequences. You'd use it as follows:

samtools view -h mapped_reads.bam | \
  samVarSplitter.pl [-ref <refName>] [-pos <int>] | \
  samtools sort > grouped_reads.bam

Then the reads can be split into different files based on their group using samtools split:

Usage: samtools split [-u <unaccounted.bam>] [-h <unaccounted_header.sam>]
                      [-f <format_string>] [-v] <merged.bam>