✈Implement new Score&Select (Single only)

pdb_profiling/data.py

@@ -0,0 +1,78 @@
+# @Created Date: 2020-10-10 04:06:02 pm
+# @Filename: data.py
+# @Email:  [email protected]
+# @Author: ZeFeng Zhu
+# @Last Modified: 2020-10-10 04:06:06 pm
+# @Copyright (c) 2020 MinghuiGroup, Soochow University
+
+blosum62 = {
+    ('W', 'F'): 1, ('L', 'R'): -2, ('S', 'P'): -1, ('V', 'T'): 0,
+    ('Q', 'Q'): 5, ('N', 'A'): -2, ('Z', 'Y'): -2, ('W', 'R'): -3,
+    ('Q', 'A'): -1, ('S', 'D'): 0, ('H', 'H'): 8, ('S', 'H'): -1,
+    ('H', 'D'): -1, ('L', 'N'): -3, ('W', 'A'): -3, ('Y', 'M'): -1,
+    ('G', 'R'): -2, ('Y', 'I'): -1, ('Y', 'E'): -2, ('B', 'Y'): -3,
+    ('Y', 'A'): -2, ('V', 'D'): -3, ('B', 'S'): 0, ('Y', 'Y'): 7,
+    ('G', 'N'): 0, ('E', 'C'): -4, ('Y', 'Q'): -1, ('Z', 'Z'): 4,
+    ('V', 'A'): 0, ('C', 'C'): 9, ('M', 'R'): -1, ('V', 'E'): -2,
+    ('T', 'N'): 0, ('P', 'P'): 7, ('V', 'I'): 3, ('V', 'S'): -2,
+    ('Z', 'P'): -1, ('V', 'M'): 1, ('T', 'F'): -2, ('V', 'Q'): -2,
+    ('K', 'K'): 5, ('P', 'D'): -1, ('I', 'H'): -3, ('I', 'D'): -3,
+    ('T', 'R'): -1, ('P', 'L'): -3, ('K', 'G'): -2, ('M', 'N'): -2,
+    ('P', 'H'): -2, ('F', 'Q'): -3, ('Z', 'G'): -2, ('X', 'L'): -1,
+    ('T', 'M'): -1, ('Z', 'C'): -3, ('X', 'H'): -1, ('D', 'R'): -2,
+    ('B', 'W'): -4, ('X', 'D'): -1, ('Z', 'K'): 1, ('F', 'A'): -2,
+    ('Z', 'W'): -3, ('F', 'E'): -3, ('D', 'N'): 1, ('B', 'K'): 0,
+    ('X', 'X'): -1, ('F', 'I'): 0, ('B', 'G'): -1, ('X', 'T'): 0,
+    ('F', 'M'): 0, ('B', 'C'): -3, ('Z', 'I'): -3, ('Z', 'V'): -2,
+    ('S', 'S'): 4, ('L', 'Q'): -2, ('W', 'E'): -3, ('Q', 'R'): 1,
+    ('N', 'N'): 6, ('W', 'M'): -1, ('Q', 'C'): -3, ('W', 'I'): -3,
+    ('S', 'C'): -1, ('L', 'A'): -1, ('S', 'G'): 0, ('L', 'E'): -3,
+    ('W', 'Q'): -2, ('H', 'G'): -2, ('S', 'K'): 0, ('Q', 'N'): 0,
+    ('N', 'R'): 0, ('H', 'C'): -3, ('Y', 'N'): -2, ('G', 'Q'): -2,
+    ('Y', 'F'): 3, ('C', 'A'): 0, ('V', 'L'): 1, ('G', 'E'): -2,
+    ('G', 'A'): 0, ('K', 'R'): 2, ('E', 'D'): 2, ('Y', 'R'): -2,
+    ('M', 'Q'): 0, ('T', 'I'): -1, ('C', 'D'): -3, ('V', 'F'): -1,
+    ('T', 'A'): 0, ('T', 'P'): -1, ('B', 'P'): -2, ('T', 'E'): -1,
+    ('V', 'N'): -3, ('P', 'G'): -2, ('M', 'A'): -1, ('K', 'H'): -1,
+    ('V', 'R'): -3, ('P', 'C'): -3, ('M', 'E'): -2, ('K', 'L'): -2,
+    ('V', 'V'): 4, ('M', 'I'): 1, ('T', 'Q'): -1, ('I', 'G'): -4,
+    ('P', 'K'): -1, ('M', 'M'): 5, ('K', 'D'): -1, ('I', 'C'): -1,
+    ('Z', 'D'): 1, ('F', 'R'): -3, ('X', 'K'): -1, ('Q', 'D'): 0,
+    ('X', 'G'): -1, ('Z', 'L'): -3, ('X', 'C'): -2, ('Z', 'H'): 0,
+    ('B', 'L'): -4, ('B', 'H'): 0, ('F', 'F'): 6, ('X', 'W'): -2,
+    ('B', 'D'): 4, ('D', 'A'): -2, ('S', 'L'): -2, ('X', 'S'): 0,
+    ('F', 'N'): -3, ('S', 'R'): -1, ('W', 'D'): -4, ('V', 'Y'): -1,
+    ('W', 'L'): -2, ('H', 'R'): 0, ('W', 'H'): -2, ('H', 'N'): 1,
+    ('W', 'T'): -2, ('T', 'T'): 5, ('S', 'F'): -2, ('W', 'P'): -4,
+    ('L', 'D'): -4, ('B', 'I'): -3, ('L', 'H'): -3, ('S', 'N'): 1,
+    ('B', 'T'): -1, ('L', 'L'): 4, ('Y', 'K'): -2, ('E', 'Q'): 2,
+    ('Y', 'G'): -3, ('Z', 'S'): 0, ('Y', 'C'): -2, ('G', 'D'): -1,
+    ('B', 'V'): -3, ('E', 'A'): -1, ('Y', 'W'): 2, ('E', 'E'): 5,
+    ('Y', 'S'): -2, ('C', 'N'): -3, ('V', 'C'): -1, ('T', 'H'): -2,
+    ('P', 'R'): -2, ('V', 'G'): -3, ('T', 'L'): -1, ('V', 'K'): -2,
+    ('K', 'Q'): 1, ('R', 'A'): -1, ('I', 'R'): -3, ('T', 'D'): -1,
+    ('P', 'F'): -4, ('I', 'N'): -3, ('K', 'I'): -3, ('M', 'D'): -3,
+    ('V', 'W'): -3, ('W', 'W'): 11, ('M', 'H'): -2, ('P', 'N'): -2,
+    ('K', 'A'): -1, ('M', 'L'): 2, ('K', 'E'): 1, ('Z', 'E'): 4,
+    ('X', 'N'): -1, ('Z', 'A'): -1, ('Z', 'M'): -1, ('X', 'F'): -1,
+    ('K', 'C'): -3, ('B', 'Q'): 0, ('X', 'B'): -1, ('B', 'M'): -3,
+    ('F', 'C'): -2, ('Z', 'Q'): 3, ('X', 'Z'): -1, ('F', 'G'): -3,
+    ('B', 'E'): 1, ('X', 'V'): -1, ('F', 'K'): -3, ('B', 'A'): -2,
+    ('X', 'R'): -1, ('D', 'D'): 6, ('W', 'G'): -2, ('Z', 'F'): -3,
+    ('S', 'Q'): 0, ('W', 'C'): -2, ('W', 'K'): -3, ('H', 'Q'): 0,
+    ('L', 'C'): -1, ('W', 'N'): -4, ('S', 'A'): 1, ('L', 'G'): -4,
+    ('W', 'S'): -3, ('S', 'E'): 0, ('H', 'E'): 0, ('S', 'I'): -2,
+    ('H', 'A'): -2, ('S', 'M'): -1, ('Y', 'L'): -1, ('Y', 'H'): 2,
+    ('Y', 'D'): -3, ('E', 'R'): 0, ('X', 'P'): -2, ('G', 'G'): 6,
+    ('G', 'C'): -3, ('E', 'N'): 0, ('Y', 'T'): -2, ('Y', 'P'): -3,
+    ('T', 'K'): -1, ('A', 'A'): 4, ('P', 'Q'): -1, ('T', 'C'): -1,
+    ('V', 'H'): -3, ('T', 'G'): -2, ('I', 'Q'): -3, ('Z', 'T'): -1,
+    ('C', 'R'): -3, ('V', 'P'): -2, ('P', 'E'): -1, ('M', 'C'): -1,
+    ('K', 'N'): 0, ('I', 'I'): 4, ('P', 'A'): -1, ('M', 'G'): -3,
+    ('T', 'S'): 1, ('I', 'E'): -3, ('P', 'M'): -2, ('M', 'K'): -1,
+    ('I', 'A'): -1, ('P', 'I'): -3, ('R', 'R'): 5, ('X', 'M'): -1,
+    ('L', 'I'): 2, ('X', 'I'): -1, ('Z', 'B'): 1, ('X', 'E'): -1,
+    ('Z', 'N'): 0, ('X', 'A'): 0, ('B', 'R'): -1, ('B', 'N'): 3,
+    ('F', 'D'): -3, ('X', 'Y'): -1, ('Z', 'R'): 0, ('F', 'H'): -1,
+    ('B', 'F'): -3, ('F', 'L'): 0, ('X', 'Q'): -1, ('B', 'B'): 4
+}

pdb_profiling/fetcher/webfetch.py

@@ -73,7 +73,7 @@ async def http_download(cls, method: str, info: Dict, path: str):
                     cls.logger.debug(f"File has been saved in: {path}")
                     return path
                 elif resp.status in (403, 404, 405):
-                    cls.logger.warning(f"403/404/405 for: {info}")
+                    cls.logger.debug(f"403/404/405 for: {info}")
                     return None
                 else:
                     mes = "code={resp.status}, message={resp.reason}, headers={resp.headers}".format(resp=resp)

pdb_profiling/pipelines/score.py

@@ -73,6 +73,9 @@ class AHP(object):
     INI_LIST = [1, 1, 2, 3, 4, 1, 2, 3, 4, 2, 3, 4, 2, 3, 2]
     factor = -1
 
+    def __init__(self, weight=None):
+        self._weight = weight
+
     @property
     def weight(self) -> List:
         com_list = list(combinations(self.ELE_LIST, 2))
@@ -93,9 +96,13 @@ def weight(self) -> List:
         return select_vector * self.factor
 
     def raw_score(self, array):
-        weight = self.weight/self.weight[0]
-        score = array[0]*weight[0]
-        return score + np.dot(array[1:], -weight[1:])
+        if self._weight is not None:
+            weight = self._weight
+        else:
+            weight = self.weight
+            weight = weight/weight[0]
+            weight[1:] = -weight[1:]
+        return np.dot(array, weight)
 
     def score(self, array):
         '''

pdb_profiling/processors/__init__.py

@@ -11,7 +11,8 @@
     PDBInterface,
     SIFTS,
     Compounds,
-    PDBs
+    PDBs,
+    SIFTSs
     )
 from pdb_profiling.processors.pdbe.api import PDBeModelServer, PDBArchive,PDBVersioned
 from pdb_profiling.processors.uniprot.api import UniProtFASTA

pdb_profiling/processors/pdbe/api.py

@@ -190,14 +190,20 @@ async def process(cls, path: Union[str, Path, Unfuture]):
         res = Dict2Tabular.pyexcel_io(traverseSuffixes(suffix, data))
         if res is not None:
             if isinstance(res, Generator):
+                one = False
                 for r in res:
-                    await pipe_out(df=r, path=new_path, format='tsv', mode='a')
+                    if r is not None:
+                        await pipe_out(df=r, path=new_path, format='tsv', mode='a')
+                        one = True
+                if not one:
+                    cls.logger.debug(f"Without Expected Data ({suffix}): {data}")
+                    return None
             else:
                 await pipe_out(df=res, path=new_path, format='tsv', mode='w')
             cls.logger.debug(f'Decoded file in {new_path}')
             return new_path
         else:
-            cls.logger.warning(f"Without Expected Data ({suffix}): {data}")
+            cls.logger.debug(f"Without Expected Data ({suffix}): {data}")
             return None
 
 
@@ -309,7 +315,7 @@ def add_tage_to_range(df: pd.DataFrame, tage_name: str):
         dfrm['unp_gaps0'] = dfrm.unp_gaps.apply(lambda x: x.count(0))
         add_tage_to_range(dfrm, tage_name='sifts_range_tag')
         dfrm['repeated'] = dfrm.apply(
-            lambda x: x['diff-'] > 0 and x['sifts_range_tag'] != 'Insertion (Specail Case)', axis=1)
+            lambda x: x['diff-'] > 0 and x['sifts_range_tag'] != 'Insertion_Undivided', axis=1)
         dfrm['repeated'] = dfrm.apply(
             lambda x: True if any(i < 0 for i in x['unp_gaps']) else x['repeated'], axis=1)
         dfrm['reversed'] = dfrm.pdb_gaps.apply(lambda x: any(i < 0 for i in x))

pdb_profiling/processors/pdbe/neo4j_api.py

@@ -26,32 +26,14 @@
 from pdb_profiling.utils import (pipe_out, sort_sub_cols, slice_series, to_interval, 
                                  lyst22intervel, SEQ_DICT, standardAA, standardNu, range_len,
                                  interval2set, lyst2range, subtract_range,
-                                 add_range, overlap_range)
+                                 add_range, overlap_range, outside_range_len)
 from pdb_profiling.log import Abclog
 from pdb_profiling.fetcher.dbfetch import Neo4j
 from pdb_profiling.processors.pdbe.sqlite_api import Sqlite_API
 import logging
 # logging.basicConfig(level=logging.INFO)
 
 
-def outside_range_len(pdb_range: Union[str, Iterable], seqres_len: int, omit: int = 5) -> int:
-    if isinstance(pdb_range, str):
-        lyst = json.loads(pdb_range)
-    else:
-        lyst = pdb_range
-    out_head = lyst[0][0]-1
-    out_tail = seqres_len - lyst[-1][-1]
-    if out_head <= omit:
-        out_head = 0
-    else:
-        out_head -= omit
-    if out_tail <= omit:
-        out_tail = 0
-    else:
-        out_tail -= omit
-    return out_head + out_tail
-
-
 def lyst2dict(lyst: List) -> Dict:
     try:
         res = dict(lyst)