initial test of masked sequence encoding (#5)

bio-seq/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "bio-seq"
-version = "0.13.0"
+version = "0.13.1"
 authors = ["jeff-k <[email protected]>"]
 edition = "2021"
 description = "Bit packed and well-typed biological sequences"

bio-seq/src/codec/masked.rs

@@ -0,0 +1,98 @@
+//! Experimental 4-bit DNA encoding with masked bases
+//! Includes `N`, `n`, `.`, `-`
+
+use crate::codec::Codec;
+use crate::seq::{ReverseComplement, Seq};
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Codec)]
+#[bits(4)]
+#[repr(u8)]
+pub enum Dna {
+    A = 0b1000,
+    C = 0b0100,
+    G = 0b0010,
+    T = 0b0001,
+
+    #[display('a')]
+    AMasked = 0b0111,
+    #[display('c')]
+    CMasked = 0b1011,
+    #[display('g')]
+    GMasked = 0b1101,
+    #[display('t')]
+    TMasked = 0b1110,
+
+    N = 0b0000,
+    #[display('n')]
+    NMasked = 0b1111,
+
+    #[display('-')]
+    #[alt(0b0011)]
+    Gap = 0b1100,
+
+    #[display('.')]
+    #[alt(0b0101)]
+    Pad = 0b1010,
+
+    #[display('?')]
+    Unknown1 = 0b0110,
+
+    #[display('!')]
+    Unknown2 = 0b1001,
+}
+
+/*
+impl Complement for Dna {
+    /// This representation can be complemented by reversing the bit pattern
+    fn comp(&self) -> Self {
+        // reverse the bits
+        Dna::unsafe_from_bits(b)
+    }
+}
+*/
+
+impl Dna {
+    /// Flipping the bit pattern masks/unmasks this representation
+    pub fn mask(&self) -> Self {
+        let b = *self as u8 ^ 0b1111;
+        Dna::unsafe_from_bits(b)
+    }
+}
+
+impl Seq<Dna> {
+    pub fn mask(&self) -> Self {
+        Self::from(!self.bv.clone())
+    }
+}
+
+impl ReverseComplement for Seq<Dna> {
+    type Output = Self;
+
+    fn revcomp(&self) -> Self {
+        let mut bv = self.bv.clone();
+        bv.reverse();
+        Self::from(bv)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::codec::masked;
+    use crate::prelude::*;
+
+    #[test]
+    fn mask_sequence() {
+        let seq = Seq::<masked::Dna>::try_from("A.TCGCgtcataN--A").unwrap();
+
+        assert_ne!(seq.mask().to_string(), "a.tcgcGTGATAN--a".to_string());
+        assert_eq!(seq.mask().to_string(), "a.tcgcGTCATAn--a".to_string());
+    }
+
+    #[test]
+    fn masked_revcomp() {
+        let seq = Seq::<masked::Dna>::try_from("A.TCGCgtcataN--A").unwrap();
+
+        assert_ne!(seq.revcomp().to_string(), "T--NtaagacGCGA.T".to_string());
+        assert_eq!(seq.revcomp().to_string(), "T--NtatgacGCGA.T".to_string());
+    }
+}

bio-seq/src/codec/mod.rs

@@ -119,6 +119,7 @@ use core::hash::Hash;
 pub mod amino;
 pub mod dna;
 pub mod iupac;
+pub mod masked;
 
 pub mod text;
 

bio-seq/src/seq/mod.rs

@@ -466,6 +466,15 @@ impl From<Vec<usize>> for Seq<text::Dna> {
     }
 }
 
+impl<A: Codec> From<Bv> for Seq<A> {
+    fn from(bv: Bv) -> Self {
+        Seq {
+            _p: PhantomData,
+            bv,
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use crate::prelude::*;