initial commit of codon-based trimming

clipkit/args_processing.py

@@ -26,6 +26,7 @@ def process_args(args) -> dict:
 
     # assign optional arguments
     complement = args.complementary or False
+    codon = args.codon or False
     mode = TrimmingMode(args.mode) if args.mode else TrimmingMode.smart_gap
     gaps = float(args.gaps) if args.gaps is not None else 0.9
     gap_characters = (
@@ -40,6 +41,7 @@ def process_args(args) -> dict:
         output_file=output_file,
         input_file_format=args.input_file_format,
         output_file_format=args.output_file_format,
+        codon=codon,
         sequence_type=sequence_type,
         complement=complement,
         gaps=gaps,

clipkit/msa.py

@@ -25,6 +25,7 @@ def __init__(
         self._site_classification_types = None
         self._column_character_frequencies = None
         self._gap_chars = gap_chars
+        self._codon_size = 3
 
     @staticmethod
     def from_bio_msa(alignment: MultipleSeqAlignment, gap_chars=None) -> "MSA":
@@ -105,6 +106,7 @@ def trim(
         mode: TrimmingMode = TrimmingMode.smart_gap,
         gap_threshold=None,
         site_positions_to_trim=None,
+        codon=False,
     ) -> np.array:
         if site_positions_to_trim is not None:
             if isinstance(site_positions_to_trim, list):
@@ -114,7 +116,9 @@ def trim(
             self._site_positions_to_trim = site_positions_to_trim
         else:
             self._site_positions_to_trim = self.determine_site_positions_to_trim(
-                mode, gap_threshold
+                mode,
+                gap_threshold,
+                codon,
             )
         self._site_positions_to_keep = np.delete(
             np.arange(self._original_length), self._site_positions_to_trim
@@ -154,7 +158,7 @@ def site_classification_types(self):
         self._site_classification_types = site_classification_types
         return self._site_classification_types
 
-    def determine_site_positions_to_trim(self, mode, gap_threshold):
+    def determine_site_positions_to_trim(self, mode, gap_threshold, codon=False):
         if mode in (TrimmingMode.gappy, TrimmingMode.smart_gap):
             sites_to_trim = np.where(self.site_gappyness >= gap_threshold)[0]
         elif mode == TrimmingMode.kpi:
@@ -200,6 +204,16 @@ def determine_site_positions_to_trim(self, mode, gap_threshold):
                 )
             )
 
+        if codon:
+            """
+            For each position in sites_to_trim we need the full triplet of codon positions tuple.
+            Example:
+                [2, 9] -> [1, 2, 3, 7, 8, 9]
+            """
+            sites_to_trim = map(
+                self.determine_codon_triplet_positions, sites_to_trim
+            ).flatten()
+            print(sites_to_trim)
         return sites_to_trim
 
     def generate_debug_log_info(self):
@@ -219,3 +233,11 @@ def generate_debug_log_info(self):
                 self.site_classification_types[idx],
                 gappyness,
             )
+
+    def determine_codon_triplet_positions(alignment_position):
+        block = alignment_position // 3
+        remainder = alignment_position % 3
+        if remainder:
+            block += 1
+        codon_triplet_index = block * 3
+        return [codon_triplet_index - 2, codon_triplet_index - 1, codon_triplet_index]

clipkit/parser.py

@@ -97,6 +97,8 @@ def create_parser() -> ArgumentParser:
         -c, --complementary                         creates complementary alignment of trimmed sequences
                                                     (input file named with '.log' suffix)
 
+        -co, --codon                                conduct trimming of codons
+
         -q, --quiet                                 disables all logging to stdout
 
         -h, --help                                  help message
@@ -151,6 +153,10 @@ def create_parser() -> ArgumentParser:
 
         Complementary
             Creates an alignment file of only the trimmed sequences
+        
+        Codon
+            Trims codon-based alignments. If one position in a codon should be trimmed, the whole
+            codon will be trimmed.
         """  # noqa
         ),
     )
@@ -255,4 +261,12 @@ def create_parser() -> ArgumentParser:
         help=SUPPRESS,
     )
 
+    optional.add_argument(
+        "-co",
+        "--codon",
+        action="store_true",
+        required=False,
+        help=SUPPRESS,
+    )
+
     return parser

tests/integration/expected/simple.fa_gappy_codon

@@ -0,0 +1,10 @@
+>1
+A-G
+>2
+A-G
+>3
+A-G
+>4
+AGA
+>5
+ACa

tests/integration/test_codon.py

@@ -0,0 +1,44 @@
+import pytest
+from pathlib import Path
+
+from clipkit.clipkit import execute
+from clipkit.files import FileFormat
+from clipkit.modes import TrimmingMode
+from clipkit.settings import DEFAULT_AA_GAP_CHARS, DEFAULT_NT_GAP_CHARS
+
+here = Path(__file__)
+
+
+@pytest.mark.integration
+class TestCodonOut(object):
+    def test_simple_codon(self):
+        """
+        test codon
+        usage: clipkit simple.fa -co
+        """
+        output_file = "output/simple.fa_gappy_codon"
+
+        kwargs = dict(
+            input_file=f"{here.parent}/samples/simple.fa",
+            output_file=output_file,
+            input_file_format="fasta",
+            output_file_format="fasta",
+            sequence_type=None,
+            complement=False,
+            codon=True,
+            gaps=0.2,
+            mode=TrimmingMode.gappy,
+            use_log=False,
+            gap_characters=DEFAULT_NT_GAP_CHARS,
+            quiet=True,
+        )
+
+        execute(**kwargs)
+
+        with open(f"{here.parent}/expected/simple.fa_gappy_codon", "r") as expected:
+            expected_content = expected.read()
+
+        with open(output_file, "r") as out_file:
+            output_content = out_file.read()
+
+        assert expected_content == output_content

tests/integration/test_complementary_output.py

@@ -26,6 +26,7 @@ def test_simple_complement(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=True,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -60,6 +61,7 @@ def test_simple_long_description_complement(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=True,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -94,6 +96,7 @@ def test_12_YIL115C_Anc_2_253_aa_aln_complement(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=True,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -128,6 +131,7 @@ def test_EOG091N44M8_aa_complement(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=True,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,

tests/integration/test_gappy_mode.py

@@ -26,6 +26,7 @@ def test_simple_no_change(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -57,6 +58,7 @@ def test_simple_no_change_long_description(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -88,6 +90,7 @@ def test_12_YIL115C_Anc_2_253_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -121,6 +124,7 @@ def test_12_YIL115C_Anc_2_253_aa_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -154,6 +158,7 @@ def test_24_ENSG00000163519_aa_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -187,6 +192,7 @@ def test_24_ENSG00000163519_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -220,6 +226,7 @@ def test_EOG091N44M8_aa(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -251,6 +258,7 @@ def test_EOG091N44M8_nt(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -316,6 +324,7 @@ def test_simple(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.2,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -349,6 +358,7 @@ def test_12_YIL115C_Anc_2_253_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.3,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -383,6 +393,7 @@ def test_24_ENSG00000163519_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.4,
             mode=TrimmingMode.gappy,
             use_log=False,
@@ -417,6 +428,7 @@ def test_EOG091N44M8_nt(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.1,
             mode=TrimmingMode.gappy,
             use_log=False,

tests/integration/test_kpi_gappy_mode.py

@@ -25,6 +25,7 @@ def test_simple(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -55,6 +56,7 @@ def test_simple_long_description(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -88,6 +90,7 @@ def test_12_YIL115C_Anc_2_253_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -122,6 +125,7 @@ def test_12_YIL115C_Anc_2_253_aa_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -155,6 +159,7 @@ def test_24_ENSG00000163519_aa_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -188,6 +193,7 @@ def test_24_ENSG00000163519_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -221,6 +227,7 @@ def test_EOG091N44M8_aa(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -254,6 +261,7 @@ def test_EOG091N44M8_nt(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.9,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -321,6 +329,7 @@ def test_simple(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.2,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -354,6 +363,7 @@ def test_12_YIL115C_Anc_2_253_codon_aln(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.3,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -388,6 +398,7 @@ def test_24_ENSG00000163519_codon_aln_custom_gap(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.4,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,
@@ -422,6 +433,7 @@ def test_EOG091N44M8_nt(self):
             output_file_format="fasta",
             sequence_type=None,
             complement=False,
+            codon=False,
             gaps=0.1,
             mode=TrimmingMode.kpi_gappy,
             use_log=False,