pandora_lib_promethion/src/vcf_reader.rs

use std::{fs::File, io::BufReader};

use csv::ReaderBuilder;
use pandora_lib_variants::variants::Variant;

use crate::{callers::Caller, variant::variant::VariantType};

#[derive(Debug, serde::Deserialize, Eq, PartialEq, Clone)]
pub struct VCFRow {
    pub chr: String,
    pub pos: u32,
    pub id: String,
    pub reference: String,
    pub alt: String,
    pub qual: String,
    pub filter: String,
    pub info: String,
    pub format: String,
    pub value: String,
}

pub fn read_vcf(
    path: &str,
    caller: &Caller,
    variant_type: &VariantType,
) -> anyhow::Result<Vec<Variant>> {
    let mut reader = ReaderBuilder::new()
        .delimiter(b'\t')
        .comment(Some(b'#'))
        .has_headers(false)
        .flexible(true)
        .from_reader(get_reader(path)?);
    let iter = reader.deserialize();

    let mut all = Vec::new();

    // should be replaced with bcftools
    for result in iter {
        let record: VCFRow = result?;

        // Normalize into multirows
        if record.alt.contains(",") {
            let alts = record.alt.split(',').collect::<Vec<&str>>();
            let n = alts.len();

            let vals = record.value.split(':').collect::<Vec<&str>>();
            let ads = vals.get(3).unwrap().split(',').collect::<Vec<&str>>();
            let vafs = vals.get(4).unwrap().split(',').collect::<Vec<&str>>();
            let pls = vals.get(5).unwrap().split(',').collect::<Vec<&str>>();

            for i in 0..n {
                let cp = &pls[(i * 3)..(i * 3 + 3)];
                let nval = format!(
                    "{}:{}:{}:{}:{}:{}",
                    vals[0],
                    vals[1],
                    vals[2],
                    [ads[0], ads[i + 1]].join(","),
                    vafs[i],
                    cp.join(",")
                );

                let mut rec = record.clone();
                rec.value = nval.clone();
                rec.alt = alts[i].to_string();

                all.push(rec);
            }
        } else {
            all.push(record);
        }
    }

    let base_n = "N".to_string();
    let res: Vec<Variant> = all
        .par_iter_mut()
        .map(|row| {
            // for Sniffles normalize insertion/deletion position (after the pos)
            if caller == &VCFSource::Sniffles && row.reference == base_n && row.alt.len() > 1 {
                row.pos -= 1;
            }
            let mut v = Variant::from_vcfrow(row, caller.clone(), variant_type.clone()).unwrap();
            v.get_depth();
            v.get_n_alt();
            v
        })
        .filter(|v| {
            for cd in v.callers_data.iter() {
                if cd.should_filter() {
                    return false;
                }
            }
            true
        })
        .collect();

    Ok(res)
}

pub fn get_reader(path: &str) -> anyhow::Result<Box<dyn std::io::Read>> {
    let file_type = *path.split(".").collect::<Vec<&str>>().last()?;
    assert!(file_type == "gz" || file_type == "vcf");

    let raw_reader: Box<dyn std::io::Read> = Box::new(File::open(path)?);

    match file_type {
        "gz" => {
            let reader = Box::new(BGZFReader::new(raw_reader)?);
            Ok(Box::new(BufReader::new(reader)))
        }
        "vcf" => {
            Ok(Box::new(BufReader::new(raw_reader)))
        }
        t => {
            panic!("unknown file type: {}", t)
        }
    }
}