geney 1.4.18__py2.py3-none-any.whl → 1.4.20__py2.py3-none-any.whl

geney/utils/SeqMats.py

@@ -30,7 +30,6 @@ def format_mut_id(text):
         return None
 
 
-
 @dataclass(slots=True)
 class SeqMat:
     """Represents a genomic sequence matrix used for training."""
@@ -43,21 +42,20 @@ class SeqMat:
     insertion_counters: dict = field(default_factory=lambda: defaultdict(int), init=False, repr=False)
     rev: bool = field(default=False, init=False, repr=False)
     predicted_splicing: pd.DataFrame = field(init=False, repr=False)
-    _pos_to_idx: dict = field(default_factory=dict, init=False, repr=False)
 
     def __init__(
-        self,
-        nucleotides: str,
-        index: np.ndarray,
-        conservation: Optional[np.ndarray] = None,
-        reference_nucleotides: Optional[np.ndarray] = None,
-        notes: Optional[dict] = None,
-        source: Optional[str] = None,
-        rev: Optional[bool] = False,
-        name: Optional[str] = 'wild_type',
-        version: Optional[str] = 'none'
+            self,
+            nucleotides: str,
+            index: np.ndarray,
+            conservation: Optional[np.ndarray] = None,
+            reference_nucleotides: Optional[np.ndarray] = None,
+            notes: Optional[dict] = None,
+            source: Optional[str] = None,
+            rev: Optional[bool] = False,
+            name: Optional[str] = 'wild_type',
+            version: Optional[str] = 'none'
     ) -> None:
-        # Initialize metadata
+        # Metadata
         self.name = name
         self.version = version
         self.source = source or "Unknown"
@@ -86,19 +84,10 @@ class SeqMat:
         self.seq_array['nt'] = nts
         self.seq_array['ref'] = nts if reference_nucleotides is None else np.array(reference_nucleotides, dtype='S1')
         self.seq_array['index'] = index
-        self.seq_array['cons'] = (np.zeros(L, dtype='f4') if conservation is None else conservation)
+        self.seq_array['cons'] = np.zeros(L, dtype='f4') if conservation is None else conservation
         self.seq_array['valid_mask'] = self.seq_array['nt'] != b'-'
 
-        # Initialize helpers
         self.insertion_counters = defaultdict(int)
-        self._build_index_map()
-
-    def _build_index_map(self):
-        """Rebuild position-to-index lookup."""
-        self._pos_to_idx = {float(pos): i for i, pos in enumerate(self.seq_array['index'])}
-
-    def __len__(self) -> int:
-        return int(self.seq_array['valid_mask'].sum())
 
     @property
     def seq(self) -> str:
@@ -114,7 +103,6 @@ class SeqMat:
 
     def clone(self, start: Optional[float] = None, end: Optional[float] = None) -> SeqMat:
         new = SeqMat.__new__(SeqMat)
-        # copy metadata
         new.name = self.name
         new.version = self.version
         new.source = self.source
@@ -123,100 +111,96 @@ class SeqMat:
         new.predicted_splicing = None
         new.insertion_counters = defaultdict(int)
 
-        # slice or full copy
         if start is not None and end is not None:
             mask = (self.seq_array['index'] >= start) & (self.seq_array['index'] <= end)
             new.seq_array = self.seq_array[mask].copy()
         else:
             new.seq_array = self.seq_array.copy()
 
-        new._build_index_map()
+        new.seq_array['valid_mask'] = new.seq_array['nt'] != b'-'
         return new
 
-    def apply_mutation(self, pos: float, ref: str, alt: str, only_snps: bool = False) -> SeqMat:
-        """Apply a single mutation to this SeqMat."""
-        # reverse-complement context
-        if self.rev:
-            self.reverse_complement()
-
-        # left-normalize
-        while ref and alt and ref[0] == alt[0]:
-            pos += 1
-            ref = ref[1:] or '-'
-            alt = alt[1:] or '-'
-
-        # substitution
-        if ref != '-' and alt != '-':
-            if len(ref) != len(alt):
-                raise ValueError("Substitution requires equal-length alleles.")
-            idx = self._pos_to_idx.get(pos)
-            if idx is None:
-                raise KeyError(f"Position {pos} not found.")
-            end = idx + len(ref)
-            if end > len(self.seq_array):
-                raise IndexError(f"Out of bounds at {pos}.")
-            # verify reference
-            ref_seg = self.seq_array['ref'][idx:end]
-            if not np.array_equal(ref_seg, np.frombuffer(ref.encode(), dtype='S1')):
-                raise ValueError(f"Ref mismatch at {pos}.")
-            # assign alt
-            self.seq_array['nt'][idx:end] = np.frombuffer(alt.encode(), dtype='S1')
-
-        # insertion
-        elif ref == '-' and alt != '-':
-            if only_snps:
-                return self
-            idx = self._pos_to_idx.get(pos)
-            if idx is None:
-                raise KeyError(f"Position {pos} not found.")
-            cnt = self.insertion_counters[pos]
-            eps = 1e-6
+    def apply_mutations(
+            self,
+            mutations: Union[Tuple[float, str, str], List[Tuple[float, str, str]]],
+            only_snps: bool = False
+    ) -> SeqMat:
+        """
+        Apply one or a batch of mutations (pos, ref, alt) efficiently:
+        - Supports a single tuple or a list of tuples
+        - Assumes mutations sorted by position for vectorized searchsorted
+        """
+        # Normalize to list
+        if isinstance(mutations, tuple) and len(mutations) == 3:
+            mutations = [mutations]
+        elif not isinstance(mutations, list):
+            raise TypeError("mutations must be a tuple or list of tuples")
+
+        # Left-normalize and bucket
+        subs, ins, dels = [], [], []
+        for pos, ref, alt in mutations:
+            while ref and alt and ref[0] == alt[0]:
+                pos += 1
+                ref = ref[1:] or '-'
+                alt = alt[1:] or '-'
+            if ref != '-' and alt != '-':
+                subs.append((pos, ref, alt))
+            elif ref == '-' and alt != '-' and not only_snps:
+                ins.append((pos, alt))
+            elif alt == '-' and ref != '-' and not only_snps:
+                dels.append((pos, ref))
+            else:
+                raise ValueError(f"Unsupported mutation {pos}:{ref}:{alt}.")
+
+        # Ensure seq_array indices sorted
+        coords = self.seq_array['index']
+
+        # 1) Bulk substitutions
+        if subs:
+            subs.sort(key=lambda x: x[0])
+            positions = np.array([p for p, _, _ in subs], dtype=coords.dtype)
+            idxs = np.searchsorted(coords, positions)
+            for (pos, ref, alt), idx in zip(subs, idxs):
+                length = len(ref)
+                if not np.all(self.seq_array['ref'][idx:idx + length] == np.frombuffer(ref.encode(), dtype='S1')):
+                    actual = self.seq_array['ref'][idx:idx + length].tobytes().decode()
+                    raise ValueError(f"Ref mismatch at {pos}: expected {ref}, found {actual}")
+                self.seq_array['nt'][idx:idx + length] = np.frombuffer(alt.encode(), dtype='S1')
+
+        # 2) Bulk insertions
+        if ins:
+            ins.sort(key=lambda x: x[0])
+            positions = np.array([p for p, _ in ins], dtype=coords.dtype)
+            idxs = np.searchsorted(coords, positions)
             new_rows = []
-            for i, nt in enumerate(alt):
-                new_rows.append((nt.encode(),
-                                 pos + (cnt + i + 1)*eps,
-                                 b'-',
-                                 np.nan,
-                                 True))
-            self._insert_rows(idx, new_rows)
-            self.insertion_counters[pos] += len(alt)
-
-        # deletion
-        elif alt == '-' and ref != '-':
-            if only_snps:
-                return self
-            idx = self._pos_to_idx.get(pos)
-            if idx is None:
-                raise KeyError(f"Position {pos} not found.")
-            end = idx + len(ref)
-            # verify
-            ref_seg = self.seq_array['ref'][idx:end]
-            if not np.array_equal(ref_seg, np.frombuffer(ref.encode(), dtype='S1')):
-                raise ValueError(f"Ref mismatch at {pos}.")
-            self.seq_array = np.delete(self.seq_array, np.s_[idx:end])
-
-        else:
-            raise ValueError("Unsupported mutation type.")
-
-        # update mask & index map
+            for (pos, alt), idx in zip(ins, idxs):
+                cnt = self.insertion_counters[pos]
+                eps = 1e-6
+                for i, nt in enumerate(alt):
+                    new_idx = pos + (cnt + i + 1) * eps
+                    new_rows.append((nt.encode(), new_idx, b'-', np.nan, True))
+                self.insertion_counters[pos] += len(alt)
+            merged = np.concatenate([self.seq_array, np.array(new_rows, dtype=self.seq_array.dtype)])
+            merged.sort(order='index')
+            self.seq_array = merged
+
+        # 3) Bulk deletions
+        if dels:
+            dels.sort(key=lambda x: x[0])
+            positions = np.array([p for p, _ in dels], dtype=coords.dtype)
+            idxs = np.searchsorted(self.seq_array['index'], positions)
+            mask = np.ones(len(self.seq_array), dtype=bool)
+            for (pos, ref), idx in zip(dels, idxs):
+                length = len(ref)
+                mask[idx:idx + length] = False
+            self.seq_array = self.seq_array[mask]
+
+        # Finalize valid mask
         self.seq_array['valid_mask'] = self.seq_array['nt'] != b'-'
-        self._build_index_map()
-
-        # restore orientation
-        if self.rev:
-            self.reverse_complement()
         return self
 
-    def _insert_rows(self, idx: int, rows: List[tuple]):
-        """Helper to insert new rows efficiently and resort."""
-        arr = self.seq_array.tolist()
-        arr[idx:idx] = rows
-        new = np.array(arr, dtype=self.seq_array.dtype)
-        new.sort(order='index')
-        self.seq_array = new
-
     def complement(self) -> SeqMat:
-        comp = {b'A':b'T', b'T':b'A', b'C':b'G', b'G':b'C', b'-':b'-'}
+        comp = {b'A': b'T', b'T': b'A', b'C': b'G', b'G': b'C', b'-': b'-'}
         nts = np.array([comp[x] for x in self.seq_array['nt']], dtype='S1')
         new = self.clone()
         new.seq_array['nt'] = nts
@@ -229,15 +213,15 @@ class SeqMat:
         return new
 
     def __getitem__(self, key: Union[int, slice]) -> np.ndarray:
-        idx = None
+        coords = self.seq_array['index']
         if isinstance(key, int):
-            idx = self._pos_to_idx.get(float(key))
-            if idx is None:
+            idx = np.searchsorted(coords, key)
+            if idx >= len(coords) or coords[idx] != key:
                 raise KeyError(f"Position {key} not found.")
             return self.seq_array[idx]
         if isinstance(key, slice):
-            start = key.start or self.min_index
-            stop = key.stop or self.max_index
-            mask = (self.seq_array['index'] >= start) & (self.seq_array['index'] <= stop)
+            start = key.start or coords.min()
+            stop = key.stop or coords.max()
+            mask = (coords >= start) & (coords <= stop)
             return self.seq_array[mask]
         raise TypeError("Invalid index type.")

{geney-1.4.18.dist-info → geney-1.4.20.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: geney
-Version: 1.4.18
+Version: 1.4.20
 Summary: A Python package for gene expression modeling.
 Home-page: https://github.com/nicolaslynn/geney
 Author: Nicolas Lynn

{geney-1.4.18.dist-info → geney-1.4.20.dist-info}/RECORD

@@ -37,7 +37,7 @@ geney/translation_initiation/tis_utils.py,sha256=AF3siFjuQH-Rs44EV-80zHdbxRMvN4w
 geney/translation_initiation/resources/kozak_pssm.json,sha256=pcd0Olziutq-6H3mFWDCD9cujQ_AlZO-iiOvBl82hqE,1165
 geney/translation_initiation/resources/tis_regressor_model.joblib,sha256=IXb4DUDhJ5rBDKcqMk9zE3ECTZZcdj7Jixz3KpoZ7OA,2592025
 geney/utils/Fasta_segment.py,sha256=weB5NJ65P0XiyAJCiCHx4T9sHC1pWLpuQeOy0B85gyg,11364
-geney/utils/SeqMats.py,sha256=2tJYPGy-cCCaANRbiYkki5yNxnhgzysDQGWjRYRSnlQ,8767
+geney/utils/SeqMats.py,sha256=q858gWPsSoS4HUr6FD1CHYuUh5AE5u9KePHYT7FQw7g,8777
 geney/utils/SeqMatsOld.py,sha256=syRU5DAuTh3xUfGW_qP9wlcBO5pHsG_y5PlrfXTIxUY,18502
 geney/utils/TranscriptLibrary.py,sha256=ma_ZVPgglxXDDneEvdqxxeqxG8eSFL-zgLUXyC6BqY8,2070
 geney/utils/__init__.py,sha256=-nJ-DMx1JzP-ZCe_QuQCeM0ZYIT_16jxoXDhUaO_4Oc,714
@@ -46,7 +46,7 @@ geney/utils/pangolin_utils.py,sha256=JQSPbWxdzqGFYfWQktkfLMaMSGR28eGQhNzO7MLMe5M
 geney/utils/spliceai_utils.py,sha256=VtrIbjyQxk_3lw86eWjftRYyal9OzxArJ0GV5u_ymTg,2721
 geney/utils/splicing_utils.py,sha256=vPCGnCPR1ooEZEHR79yFHLmRQXEJHXEQjjxpBR-YWOs,20635
 geney/utils/utils.py,sha256=m51Vd0cEbrcIHo6_8BAuI9YSPcKRs22e5LfVd2Qj6Is,2181
-geney-1.4.18.dist-info/METADATA,sha256=WLdB6CQyOBWNDYLU05Gyj6o7kaNzA5qr7ROD0mzTgm4,990
-geney-1.4.18.dist-info/WHEEL,sha256=AHX6tWk3qWuce7vKLrj7lnulVHEdWoltgauo8bgCXgU,109
-geney-1.4.18.dist-info/top_level.txt,sha256=O-FuNUMb5fn9dhZ-dYCgF0aZtfi1EslMstnzhc5IIVo,6
-geney-1.4.18.dist-info/RECORD,,
+geney-1.4.20.dist-info/METADATA,sha256=NznExaPOYo1PvAj3-Uxq3Y9SXMNLd__iz30PKF6ix18,990
+geney-1.4.20.dist-info/WHEEL,sha256=AHX6tWk3qWuce7vKLrj7lnulVHEdWoltgauo8bgCXgU,109
+geney-1.4.20.dist-info/top_level.txt,sha256=O-FuNUMb5fn9dhZ-dYCgF0aZtfi1EslMstnzhc5IIVo,6
+geney-1.4.20.dist-info/RECORD,,