feat: allow importing gnomAD SV/CNV v4 as background db (#295)

src/strucvars/txt_to_bin/cli.rs

@@ -72,8 +72,12 @@ pub enum InputType {
     StrucvarExacCnv,
     /// Convert Thousand Genomes to binary.
     StrucvarG1k,
-    /// Convert gnomAD SV to binary.
-    StrucvarGnomadSv,
+    /// Convert gnomAD SV v2 to binary.
+    StrucvarGnomadSv2,
+    /// Convert gnomAD CNV v4 to binary.
+    StrucvarGnomadCnv4,
+    /// Convert gnomAD SV v4 to binary.
+    StrucvarGnomadSv4,
     /// Convert masked region to binary.
     MaskedRegion,
     /// Convert cross-link to binary.
@@ -100,7 +104,7 @@ pub struct Args {
 
 /// Main entry point for the `strucvars txt-to-bin` command.
 pub fn run(common_args: &crate::common::Args, args: &Args) -> Result<(), anyhow::Error> {
-    tracing::info!("Starting `db to-bin`");
+    tracing::info!("Starting `strucvars txt-to-bin`");
     tracing::info!("  common_args = {:?}", &common_args);
     tracing::info!("  args = {:?}", &args);
 
@@ -135,9 +139,21 @@ pub fn run(common_args: &crate::common::Args, args: &Args) -> Result<(), anyhow:
         InputType::StrucvarG1k => {
             vardbs::convert_to_bin(&args.path_input, &args.path_output, InputFileType::G1k)?
         }
-        InputType::StrucvarGnomadSv => {
-            vardbs::convert_to_bin(&args.path_input, &args.path_output, InputFileType::Gnomad)?
-        }
+        InputType::StrucvarGnomadSv2 => vardbs::convert_to_bin(
+            &args.path_input,
+            &args.path_output,
+            InputFileType::GnomadSv2,
+        )?,
+        InputType::StrucvarGnomadCnv4 => vardbs::convert_to_bin(
+            &args.path_input,
+            &args.path_output,
+            InputFileType::GnomadCnv4,
+        )?,
+        InputType::StrucvarGnomadSv4 => vardbs::convert_to_bin(
+            &args.path_input,
+            &args.path_output,
+            InputFileType::GnomadSv4,
+        )?,
         InputType::MaskedRegion => masked::convert_to_bin(&args.path_input, &args.path_output)?,
         InputType::Xlink => xlink::convert_to_bin(&args.path_input, &args.path_output)?,
     }
@@ -301,14 +317,14 @@ mod test {
     }
 
     #[test]
-    fn run_strucvar_gnomad_smoke() -> Result<(), anyhow::Error> {
+    fn run_strucvar_gnomad_sv2_smoke() -> Result<(), anyhow::Error> {
         let tmp_dir = temp_testdir::TempDir::default();
         let common_args = common::Args {
             verbose: Verbosity::new(0, 0),
         };
         let args = Args {
             assembly: None,
-            input_type: InputType::StrucvarGnomadSv,
+            input_type: InputType::StrucvarGnomadSv2,
             path_input: String::from(
                 "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/gnomad_sv.bed.gz",
             ),
@@ -320,6 +336,46 @@ mod test {
         Ok(())
     }
 
+    #[test]
+    fn run_strucvar_gnomad_cnv4_smoke() -> Result<(), anyhow::Error> {
+        let tmp_dir = temp_testdir::TempDir::default();
+        let common_args = common::Args {
+            verbose: Verbosity::new(0, 0),
+        };
+        let args = Args {
+            assembly: None,
+            input_type: InputType::StrucvarGnomadCnv4,
+            path_input: String::from(
+                "tests/db/to-bin/varfish-db-downloader/vardbs/grch38/strucvar/gnomad-cnv.bed.gz",
+            ),
+            path_output: tmp_dir.join("gnomad-cnv.bin"),
+        };
+
+        super::run(&common_args, &args)?;
+
+        Ok(())
+    }
+
+    #[test]
+    fn run_strucvar_gnomad_sv4_smoke() -> Result<(), anyhow::Error> {
+        let tmp_dir = temp_testdir::TempDir::default();
+        let common_args = common::Args {
+            verbose: Verbosity::new(0, 0),
+        };
+        let args = Args {
+            assembly: None,
+            input_type: InputType::StrucvarGnomadSv4,
+            path_input: String::from(
+                "tests/db/to-bin/varfish-db-downloader/vardbs/grch38/strucvar/gnomad-sv.bed.gz",
+            ),
+            path_output: tmp_dir.join("gnomad-sv.bin"),
+        };
+
+        super::run(&common_args, &args)?;
+
+        Ok(())
+    }
+
     #[test]
     fn run_masked_region_smoke() -> Result<(), anyhow::Error> {
         let tmp_dir = temp_testdir::TempDir::default();

src/strucvars/txt_to_bin/vardbs/input.rs

@@ -85,9 +85,9 @@ pub struct G1kRecord {
     pub n_het: u32,
 }
 
-/// gnomAD SV database record as read from TSV file.
+/// gnomAD SV v2 database record as read from TSV file.
 #[derive(Debug, Deserialize)]
-pub struct GnomadRecord {
+pub struct GnomadSv2Record {
     /// chromosome name
     pub chromosome: String,
     /// begin position, 0-based
@@ -102,6 +102,56 @@ pub struct GnomadRecord {
     pub n_het: u32,
 }
 
+/// gnomAD SV v4 database record as read from TSV file.
+#[derive(Debug, Deserialize)]
+pub struct GnomadSv4Record {
+    /// chromosome name
+    pub chromosome: String,
+    /// begin position, 0-based
+    pub begin: i32,
+    /// end position, 0-based
+    pub end: i32,
+    /// The structural vairant type
+    pub svtype: String,
+    /// Number of male homozygous reference allele carriers.
+    pub male_n_homref: u32,
+    /// Number of male heterozygous alternate allele carriers.
+    pub male_n_het: u32,
+    /// Number of male homozygous alternate allele carriers.
+    pub male_n_homalt: u32,
+    /// Number of male hemizygous alternate allele carriers.
+    pub male_n_hemiref: u32,
+    /// Number of male hemizygous reference allele carriers.
+    pub male_n_hemialt: u32,
+    /// Number of female homozygous reference allele carriers.
+    pub female_n_homref: u32,
+    /// Number of female heterozygous alternate allele carriers.
+    pub female_n_het: u32,
+    /// Number of female homozygous alternate allele carriers.
+    pub female_n_homalt: u32,
+    /// Number of samples at this site (CNV only).
+    pub cnv_n_total: u32,
+    /// Number of samples with a CNV at this site (CNV only).
+    pub cnv_n_var: u32,
+}
+
+/// gnomAD CNV v$ database record as read from TSV file.
+#[derive(Debug, Deserialize)]
+pub struct GnomadCnv4Record {
+    /// chromosome name
+    pub chromosome: String,
+    /// begin position, 0-based
+    pub begin: i32,
+    /// end position, 0-based
+    pub end: i32,
+    /// The structural vairant type
+    pub svtype: String,
+    /// Number of samples at this site (passing QC).
+    pub n_total: u32,
+    /// Number of samples with a CNV at this site (passing QC).
+    pub n_var: u32,
+}
+
 /// Common type to convert input data to.
 pub struct InputRecord {
     /// Chromosome of start position.
@@ -251,7 +301,7 @@ impl TryInto<Option<InputRecord>> for ExacRecord {
     }
 }
 
-impl TryInto<Option<InputRecord>> for GnomadRecord {
+impl TryInto<Option<InputRecord>> for GnomadSv2Record {
     type Error = &'static str;
 
     fn try_into(self) -> Result<Option<InputRecord>, Self::Error> {
@@ -279,6 +329,59 @@ impl TryInto<Option<InputRecord>> for GnomadRecord {
     }
 }
 
+impl TryInto<Option<InputRecord>> for GnomadCnv4Record {
+    type Error = &'static str;
+
+    fn try_into(self) -> Result<Option<InputRecord>, Self::Error> {
+        Ok(Some(InputRecord {
+            chromosome: self.chromosome.clone(),
+            chromosome2: self.chromosome,
+            begin: self.begin,
+            end: self.end,
+            sv_type: match self.svtype.as_str() {
+                "DEL" => SvType::Del,
+                "DUP" => SvType::Dup,
+                _ => {
+                    error!("sv_type = {}", &self.svtype);
+                    return Err("unknown SV type");
+                }
+            },
+            count: self.n_var,
+        }))
+    }
+}
+
+impl TryInto<Option<InputRecord>> for GnomadSv4Record {
+    type Error = &'static str;
+
+    fn try_into(self) -> Result<Option<InputRecord>, Self::Error> {
+        Ok(Some(InputRecord {
+            chromosome: self.chromosome.clone(),
+            chromosome2: self.chromosome,
+            begin: self.begin,
+            end: self.end,
+            sv_type: match self.svtype.as_str() {
+                "BND" => SvType::Bnd,
+                "CNV" => SvType::Cnv,
+                "DEL" => SvType::Del,
+                "DUP" => SvType::Dup,
+                "INS" => SvType::Ins,
+                "INV" => SvType::Inv,
+                _ => {
+                    error!("sv_type = {}", &self.svtype);
+                    return Err("unknown SV type");
+                }
+            },
+            count: self.male_n_het
+                + self.male_n_homalt
+                + self.male_n_hemialt
+                + self.female_n_het
+                + self.female_n_homalt
+                + self.cnv_n_var,
+        }))
+    }
+}
+
 impl TryInto<Option<InputRecord>> for G1kRecord {
     type Error = &'static str;
 

src/strucvars/txt_to_bin/vardbs/mod.rs

@@ -27,7 +27,9 @@ pub enum InputFileType {
     DgvGs,
     Exac,
     G1k,
-    Gnomad,
+    GnomadSv2,
+    GnomadCnv4,
+    GnomadSv4,
     InhouseDb,
 }
 /// Deserialize from CSV reader to an `Option<records::InputRecord>`
@@ -76,6 +78,24 @@ where
     Ok(result)
 }
 
+/// Branch around `deserialize_loop`.
+pub fn deserialize_branch(
+    input_type: InputFileType,
+    reader: &mut csv::Reader<Box<dyn std::io::BufRead>>,
+) -> Result<Vec<BgDbRecord>, anyhow::Error> {
+    match input_type {
+        InputFileType::Dbvar => deserialize_loop::<input::DbVarRecord>(reader),
+        InputFileType::Dgv => deserialize_loop::<input::DgvRecord>(reader),
+        InputFileType::DgvGs => deserialize_loop::<input::DgvGsRecord>(reader),
+        InputFileType::Exac => deserialize_loop::<input::ExacRecord>(reader),
+        InputFileType::G1k => deserialize_loop::<input::G1kRecord>(reader),
+        InputFileType::InhouseDb => deserialize_loop::<InhouseDbRecord>(reader),
+        InputFileType::GnomadSv2 => deserialize_loop::<input::GnomadSv2Record>(reader),
+        InputFileType::GnomadCnv4 => deserialize_loop::<input::GnomadCnv4Record>(reader),
+        InputFileType::GnomadSv4 => deserialize_loop::<input::GnomadSv4Record>(reader),
+    }
+}
+
 /// Perform conversion to protobuf `.bin` file.
 pub fn convert_to_bin<P, Q>(
     path_input_tsv: P,
@@ -97,15 +117,7 @@ where
         )?);
     let before_parsing = Instant::now();
 
-    let records = match input_type {
-        InputFileType::Dbvar => deserialize_loop::<input::DbVarRecord>(&mut reader)?,
-        InputFileType::Dgv => deserialize_loop::<input::DgvRecord>(&mut reader)?,
-        InputFileType::DgvGs => deserialize_loop::<input::DgvGsRecord>(&mut reader)?,
-        InputFileType::Exac => deserialize_loop::<input::ExacRecord>(&mut reader)?,
-        InputFileType::G1k => deserialize_loop::<input::G1kRecord>(&mut reader)?,
-        InputFileType::Gnomad => deserialize_loop::<input::GnomadRecord>(&mut reader)?,
-        InputFileType::InhouseDb => deserialize_loop::<InhouseDbRecord>(&mut reader)?,
-    };
+    let records = deserialize_branch(input_type, &mut reader)?;
     let bg_db = BackgroundDatabase { records };
 
     tracing::debug!(
@@ -127,3 +139,73 @@ where
 
     Ok(())
 }
+
+#[cfg(test)]
+mod test {
+    use super::InputFileType;
+
+    #[rstest::rstest]
+    #[case::dbvar(
+        InputFileType::Dbvar,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/dbvar.bed.gz"
+    )]
+    #[case::dgv(
+        InputFileType::Dgv,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/dgv.bed.gz"
+    )]
+    #[case::dgv_gs(
+        InputFileType::DgvGs,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/dgv_gs.bed.gz"
+    )]
+    #[case::exac(
+        InputFileType::Exac,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/exac.bed.gz"
+    )]
+    #[case::g1k(
+        InputFileType::G1k,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/g1k.bed.gz"
+    )]
+    #[case::gnomad_sv2(
+        InputFileType::GnomadSv2,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/gnomad_sv.bed.gz"
+    )]
+    #[case::gnomad_cnv4(
+        InputFileType::GnomadCnv4,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch38/strucvar/gnomad-cnv.bed.gz"
+    )]
+    #[case::gnomad_sv4(
+        InputFileType::GnomadSv4,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch38/strucvar/gnomad-sv.bed.gz"
+    )]
+    #[case::inhouse_db(
+        InputFileType::InhouseDb,
+        "tests/db/to-bin/varfish-db-downloader/vardbs/grch37/strucvar/inhouse.tsv"
+    )]
+    fn test_deserialize_branch(
+        #[case] input_type: InputFileType,
+        #[case] path_input: &str,
+    ) -> Result<(), anyhow::Error> {
+        mehari::common::set_snapshot_suffix!(
+            "{:?}-{}",
+            input_type,
+            path_input
+                .split('/')
+                .last()
+                .unwrap()
+                .split('.')
+                .next()
+                .unwrap()
+        );
+
+        let mut reader = csv::ReaderBuilder::new()
+            .has_headers(false)
+            .comment(Some(b'#'))
+            .delimiter(b'\t')
+            .from_reader(mehari::common::io::std::open_read_maybe_gz(path_input)?);
+
+        let records = super::deserialize_branch(input_type, &mut reader)?;
+        insta::assert_yaml_snapshot!(records);
+
+        Ok(())
+    }
+}