geney 1.3.2__py2.py3-none-any.whl → 1.3.4__py2.py3-none-any.whl

geney/oncosplice.py

@@ -6,6 +6,10 @@ import pandas as pd
 import numpy as np
 from .SeqMats import SeqMat, MutSeqMat
 from .splicing_utils import find_transcript_missplicing_seqs, develop_aberrant_splicing
+from .Gene import Gene
+import copy
+from . import config
+
 from .tis_utils import find_tis
 
 def short_hash_of_list(numbers, length=5):
@@ -301,7 +305,7 @@ def OncospliceAnnotator(reference_transcript, variant_transcript, mut, ref_attri
 
 
 def oncosplice(mut_id, splicing_threshold=0.5, protein_coding=True, cons_required=False, primary_transcript=False,
-               window_length=13, organism='hg38', engine='spliceai', domains=None):
+               window_length=13, organism='hg38', engine='spliceai'):
     gene = Gene.from_file(mut_id.split(':')[0], organism=organism)
     reference_gene_proteins = {
         transcript.generate_pre_mrna().generate_mature_mrna().generate_protein().protein: transcript.transcript_id for
@@ -374,110 +378,110 @@ def oncosplice(mut_id, splicing_threshold=0.5, protein_coding=True, cons_require
          'mutation_distance_from_3', 'engine', 'reference_resemblance', 'oncosplice_score', 'percentile',
          'isoform_prevalence', 'reference_protein', 'variant_protein']]
 
-
-import asyncio
-async def oncosplice_prototype(mut_id, splicing_threshold=0.5, protein_coding=True, primary_transcript=False,
-                               window_length=13, organism='hg38', engine='spliceai', use_cons=True, require_cons=False):
-    import sys, os
-    needed_file1 = config[organism]['yoram_path'] / 'rest_api_utils.py'
-    needed_file2 = config[organism]['yoram_path'] / 'uniprot_utils.py'
-
-    if sys.platform == 'linux' and (needed_file1.is_file() and os.access(needed_file1, os.R_OK)) and (
-            needed_file2.is_file() and os.access(needed_file2, os.R_OK)):
-        sys.path.append(str(config[organism]['yoram_path']))
-        import uniprot_utils as uput
-
-    else:
-        raise SystemError(
-            "Oncosplice Prototype can only be run on Power with access to the /tamir2/yoramzar/Projects/Cancer_mut/Utils folder.")
-
-    from .tis_utils import find_tis
-
-    # Define async functions
-    async def background_request(ensb_id, Uniprot_features=["Topological domain", "Transmembrane", "Domain"]):
-        return uput.retrieve_protein_data_features_subset(uput.ensembl_id2uniprot_id(ensb_id), Uniprot_features)
-
-    def inspect_domain(row, modified_vector, conservation_vector):
-        v1, v2 = modified_vector[row.start:row.end], conservation_vector[row.start:row.end]
-        if sum(v2) == 0:
-            return pd.Series([f'{row.type}|{row.start}|{row.end}|{row.description}', 0],
-                             index=['domain_identifier', 'score'])
-
-        return pd.Series([f'{row.type}|{row.start}|{row.end}|{row.description}', sum(v1 * v2) / sum(v2)],
-                         index=['domain_identifier', 'score'])
-
-    gene = Gene(mut_id.split(':')[0], organism=organism)
-    reference_gene_proteins = {tid: transcript.generate_pre_mrna().generate_mature_mrna().generate_protein() for tid, transcript in gene.run_transcripts(protein_coding=True)}
-    mutations = [get_mutation(mut_id, rev=gene.rev) for mut_id in mut_id.split('|')]
-    results = []
-    for tid, transcript in gene.run_transcripts(protein_coding=protein_coding, primary_transcript=primary_transcript):
-        if require_cons and not transcript.cons_available:
-            continue
-
-        if all(mutation not in transcript for mutation in mutations):
-            # results.append({'transcript_id': transcript.transcript_id})
-            continue
-
-        task1 = asyncio.create_task(background_request(tid))
-        transcript.generate_pre_mrna()
-        transcript.cons_vector = transform_conservation_vector(transcript.cons_vector, window=window_length)
-        transcript.generate_mature_mrna().generate_protein(inplace=True)
-        ref_protein, cons_vector = transcript.protein, transcript.cons_vector
-
-        if not use_cons:
-            cons_vector = np.ones(len(ref_protein))
-
-        if sum(cons_vector) == 0:
-            cons_vector = np.ones(len(ref_protein)) #/len(ref_protein)
-
-        reference_transcript = copy.deepcopy(transcript)
-
-        assert len(ref_protein) == len(
-            cons_vector), f"Protein ({len(ref_protein)}) and conservation vector ({len(cons_vector)} must be same length."
-
-        missplicing = Missplicing(find_transcript_missplicing(transcript, mutations, engine=engine, threshold=splicing_threshold),
-                                  threshold=splicing_threshold)
-        for mutation in mutations:
-            transcript.pre_mrna += mutation
-
-        domains_df = await task1
-        for i, new_boundaries in enumerate(develop_aberrant_splicing(transcript, missplicing.aberrant_splicing)):
-            transcript.acceptors = new_boundaries['acceptors']
-            transcript.donors = new_boundaries['donors']
-            transcript.generate_mature_mrna()
-            transcript.TIS = find_tis(ref_seq=reference_transcript, mut_seq=transcript)
-            transcript.generate_protein()
-
-            alignment = get_logical_alignment(reference_transcript.protein, transcript.protein)
-            deleted, inserted = find_indels_with_mismatches_as_deletions(alignment.seqA, alignment.seqB)
-            modified_positions = find_modified_positions(len(ref_protein), deleted, inserted)
-            temp_cons = np.convolve(cons_vector * modified_positions, np.ones(window_length)) / window_length
-            affected_cons_scores = max(temp_cons)
-            percentile = (
-                    sorted(cons_vector).index(next(x for x in sorted(cons_vector) if x >= affected_cons_scores)) / len(
-                cons_vector))
-
-            out = domains_df.apply(lambda row: inspect_domain(row, modified_positions, cons_vector), axis=1)
-            domains_affected = '+'.join([f'{a}:{round(b, 3)}' for a, b in list(zip(out.domain_identifier, out.score))])
-
-            report = OncospliceAnnotator(reference_transcript, transcript, mutation)
-            report['mut_id'] = mut_id
-            report['oncosplice_score'] = affected_cons_scores
-            report['cons_available'] = transcript.cons_available
-            report['transcript_id'] = transcript.transcript_id
-            report['percentile'] = percentile
-            report['isoform_id'] = i
-            report['isoform_prevalence'] = new_boundaries['path_weight']
-            report['full_missplicing'] = missplicing.aberrant_splicing
-            report['missplicing'] = max(missplicing)
-            report['domains'] = domains_affected
-            report['max_domain_score'] = out.score.max()
-
-            report['reference_resemblance'] = reference_gene_proteins.get(transcript.protein, None)
-            results.append(pd.Series(report))
-
-    report = pd.concat(results, axis=1).T
-    return report
+#
+# import asyncio
+# async def oncosplice_prototype(mut_id, splicing_threshold=0.5, protein_coding=True, primary_transcript=False,
+#                                window_length=13, organism='hg38', engine='spliceai', use_cons=True, require_cons=False):
+#     import sys, os
+#     needed_file1 = config[organism]['yoram_path'] / 'rest_api_utils.py'
+#     needed_file2 = config[organism]['yoram_path'] / 'uniprot_utils.py'
+#
+#     if sys.platform == 'linux' and (needed_file1.is_file() and os.access(needed_file1, os.R_OK)) and (
+#             needed_file2.is_file() and os.access(needed_file2, os.R_OK)):
+#         sys.path.append(str(config[organism]['yoram_path']))
+#         import uniprot_utils as uput
+#
+#     else:
+#         raise SystemError(
+#             "Oncosplice Prototype can only be run on Power with access to the /tamir2/yoramzar/Projects/Cancer_mut/Utils folder.")
+#
+#     from .tis_utils import find_tis
+#
+#     # Define async functions
+#     async def background_request(ensb_id, Uniprot_features=["Topological domain", "Transmembrane", "Domain"]):
+#         return uput.retrieve_protein_data_features_subset(uput.ensembl_id2uniprot_id(ensb_id), Uniprot_features)
+#
+#     def inspect_domain(row, modified_vector, conservation_vector):
+#         v1, v2 = modified_vector[row.start:row.end], conservation_vector[row.start:row.end]
+#         if sum(v2) == 0:
+#             return pd.Series([f'{row.type}|{row.start}|{row.end}|{row.description}', 0],
+#                              index=['domain_identifier', 'score'])
+#
+#         return pd.Series([f'{row.type}|{row.start}|{row.end}|{row.description}', sum(v1 * v2) / sum(v2)],
+#                          index=['domain_identifier', 'score'])
+#
+#     gene = Gene(mut_id.split(':')[0], organism=organism)
+#     reference_gene_proteins = {tid: transcript.generate_pre_mrna().generate_mature_mrna().generate_protein() for tid, transcript in gene.run_transcripts(protein_coding=True)}
+#     mutations = [get_mutation(mut_id, rev=gene.rev) for mut_id in mut_id.split('|')]
+#     results = []
+#     for tid, transcript in gene.run_transcripts(protein_coding=protein_coding, primary_transcript=primary_transcript):
+#         if require_cons and not transcript.cons_available:
+#             continue
+#
+#         if all(mutation not in transcript for mutation in mutations):
+#             # results.append({'transcript_id': transcript.transcript_id})
+#             continue
+#
+#         task1 = asyncio.create_task(background_request(tid))
+#         transcript.generate_pre_mrna()
+#         transcript.cons_vector = transform_conservation_vector(transcript.cons_vector, window=window_length)
+#         transcript.generate_mature_mrna().generate_protein(inplace=True)
+#         ref_protein, cons_vector = transcript.protein, transcript.cons_vector
+#
+#         if not use_cons:
+#             cons_vector = np.ones(len(ref_protein))
+#
+#         if sum(cons_vector) == 0:
+#             cons_vector = np.ones(len(ref_protein)) #/len(ref_protein)
+#
+#         reference_transcript = copy.deepcopy(transcript)
+#
+#         assert len(ref_protein) == len(
+#             cons_vector), f"Protein ({len(ref_protein)}) and conservation vector ({len(cons_vector)} must be same length."
+#
+#         missplicing = Missplicing(find_transcript_missplicing(transcript, mutations, engine=engine, threshold=splicing_threshold),
+#                                   threshold=splicing_threshold)
+#         for mutation in mutations:
+#             transcript.pre_mrna += mutation
+#
+#         domains_df = await task1
+#         for i, new_boundaries in enumerate(develop_aberrant_splicing(transcript, missplicing.aberrant_splicing)):
+#             transcript.acceptors = new_boundaries['acceptors']
+#             transcript.donors = new_boundaries['donors']
+#             transcript.generate_mature_mrna()
+#             transcript.TIS = find_tis(ref_seq=reference_transcript, mut_seq=transcript)
+#             transcript.generate_protein()
+#
+#             alignment = get_logical_alignment(reference_transcript.protein, transcript.protein)
+#             deleted, inserted = find_indels_with_mismatches_as_deletions(alignment.seqA, alignment.seqB)
+#             modified_positions = find_modified_positions(len(ref_protein), deleted, inserted)
+#             temp_cons = np.convolve(cons_vector * modified_positions, np.ones(window_length)) / window_length
+#             affected_cons_scores = max(temp_cons)
+#             percentile = (
+#                     sorted(cons_vector).index(next(x for x in sorted(cons_vector) if x >= affected_cons_scores)) / len(
+#                 cons_vector))
+#
+#             out = domains_df.apply(lambda row: inspect_domain(row, modified_positions, cons_vector), axis=1)
+#             domains_affected = '+'.join([f'{a}:{round(b, 3)}' for a, b in list(zip(out.domain_identifier, out.score))])
+#
+#             report = OncospliceAnnotator(reference_transcript, transcript, mutation)
+#             report['mut_id'] = mut_id
+#             report['oncosplice_score'] = affected_cons_scores
+#             report['cons_available'] = transcript.cons_available
+#             report['transcript_id'] = transcript.transcript_id
+#             report['percentile'] = percentile
+#             report['isoform_id'] = i
+#             report['isoform_prevalence'] = new_boundaries['path_weight']
+#             report['full_missplicing'] = missplicing.aberrant_splicing
+#             report['missplicing'] = max(missplicing)
+#             report['domains'] = domains_affected
+#             report['max_domain_score'] = out.score.max()
+#
+#             report['reference_resemblance'] = reference_gene_proteins.get(transcript.protein, None)
+#             results.append(pd.Series(report))
+#
+#     report = pd.concat(results, axis=1).T
+#     return report
 
 
 if __name__ == '__main__':

geney/splicing_utils.py

@@ -1,7 +1,6 @@
 import numpy as np
-from .mutation_utils import get_mutation
-from .seqmat_utils import Gene
-
+from .Gene import Gene
+from .SeqMats import MutSeqMat
 from collections import defaultdict
 
 
@@ -145,75 +144,185 @@ def find_ss_changes(ref_dct, mut_dct, known_splice_sites, threshold=0.5):
     return discovered_pos, deleted_pos
 
 
-def find_transcript_missplicing_mutid(mut_id):
-    from geney.Gene import Gene
-    transcript = Gene(mut_id.split(':')[0]).transcript().generate_mature_mrna()
-    out = find_transcript_missplicing(transcript, [get_mutation(mut_id, rev=transcript.rev)], context=5000, window=2500, threshold=0.5, engine='spliceai', just_ss=True)
-    best_delta = 0
-    for k, v in out.items():
-        for k1, v1 in v.items():
-            if abs(v1['delta']) > abs(best_delta):
-                best_delta = v1['delta']
-    return out, best_delta
-
-def find_transcript_missplicing(transcript, mutations, context=5000, window=2500, threshold=0.5, engine='spliceai', just_ss=False):
-    from functools import reduce
-    ref = transcript.pre_mrna
-    mutations = [mutation for mutation in mutations if mutation.position in ref.indices]
-    if len(mutations) == 0:
-        return {'missed_acceptors': {}, 'missed_donors': {}, 'discovered_acceptors': {}, 'discovered_donors': {}}
+from typing import Tuple, Dict
+
+def run_splicing_engine(seq, engine='spliceai'):
+    match engine:
+        case 'spliceai':
+            from .spliceai_utils import sai_predict_probs, sai_models
+            donor_probs, acceptor_probs = sai_predict_probs(seq, models=sai_models)
+
+        case 'pangolin':
+            from .pangolin_utils import pangolin_predict_probs, pang_models
+            donor_probs, acceptor_probs = pangolin_predict_probs(seq, models=pang_models)
+
+        case _:
+            raise ValueError(f"{engine} not implemented")
+    return donor_probs, acceptor_probs
 
-    var = reduce(lambda acc, mutation: acc + mutation, mutations, ref)
-    center = int(np.mean([mutation.position for mutation in mutations]) // 1)
 
-    total_context = context + window
-    length = ref.seqmat.shape[-1]
-    center_index = ref.rel_pos(center)
-    ref_start_pad = max(0, total_context - center_index)
-    ref_end_pad = max(0, total_context - (length - center_index))
+def find_transcript_splicing(transcript, engine: str = 'spliceai') -> Tuple[Dict[int, float], Dict[int, float]]:
+    """
+    Predict splice site probabilities for a given transcript using the specified engine.
+    This function uses a padding of 5000 'N's on each side of the transcript sequence
+    to align with the model's required context length.
+
+    Args:
+        transcript: An object representing a transcript, expected to have:
+            - an `indices` attribute that returns a sequence of positions.
+            - a `seq` attribute that returns the sequence string.
+        engine (str): The prediction engine to use. Supported: 'spliceai', 'pangolin'.
+
+    Returns:
+        (donor_probs, acceptor_probs) as two dictionaries keyed by position with probability values.
+
+    Raises:
+        ValueError: If an unsupported engine is provided.
+        AssertionError: If the length of predicted probabilities does not match the length of indices.
+    """
+    # Prepare reference sequence with padding
+    ref_indices = transcript.indices
+    ref_seq = 'N' * 5000 + transcript.seq + 'N' * 5000
+    ref_seq_acceptor_probs, ref_seq_donor_probs = run_splicing_engine(ref_seq, engine)
+
+    # Verify lengths
+    assert len(ref_seq_donor_probs) == len(ref_indices), (
+        f"Donor probabilities length ({len(ref_seq_donor_probs)}) does not match "
+        f"indices length ({len(ref_indices)})."
+    )
+    assert len(ref_seq_acceptor_probs) == len(ref_indices), (
+        f"Acceptor probabilities length ({len(ref_seq_acceptor_probs)}) does not match "
+        f"indices length ({len(ref_indices)})."
+    )
+
+    # Create dictionaries and sort them by probability in descending order
+    donor_probs = dict(sorted((i, p) for i, p in zip(ref_indices, ref_seq_donor_probs)),
+                       key=lambda item: item[1], reverse=True)
+    acceptor_probs = dict(sorted((i, p) for i, p in zip(ref_indices, ref_seq_acceptor_probs)),
+                          key=lambda item: item[1], reverse=True)
+
+    return donor_probs, acceptor_probs
 
-    length = var.seqmat.shape[-1]
-    center_index = var.rel_pos(center)
-    if center_index is None:
-        raise IndexError("Center index must not be none... Issue with mutations... They must not be within the transcript.")
-    var_start_pad = max(0, total_context - center_index)
-    var_end_pad = max(0, total_context - (length - center_index))
 
-    ref = ref.inspect(center, context=total_context)
-    var = var.inspect(center, context=total_context)
+def find_transcript_missplicing(mut_id, transcript='primary', threshold=0.5, engine='spliceai', organism='hg38'):
+    gene = Gene.from_file(mut_id.split(':')[0], organism=organism)
+    reference_transcript = gene.transcript(transcript) if transcript is not None else gene.transcript()
+    variant_transcript = reference_transcript.clone()
+    mutations = [MutSeqMat.from_mutid(m) for m in mut_id.split('|')]
+    mutations = [m for m in mutations if m in reference_transcript]
+    if len(mutations) == 0:
+        return {'missed_acceptors': {}, 'missed_donors': {}, 'discovered_acceptors': {}, 'discovered_donors': {}}
 
-    ref_indices = np.concatenate([np.zeros(ref_start_pad), ref.indices, np.zeros(ref_end_pad)])
-    mut_indices = np.concatenate([np.zeros(var_start_pad),  var.indices, np.zeros(var_end_pad)])
+    center = np.mean([m.indices[0] for m in mutations]) // 1
+    for mutation in mutations:
+        variant_transcript.mutate(mutation, inplace=True)
 
-    ref_indices = ref_indices[context:-context]
-    mut_indices = mut_indices[context:-context]
+    return find_transcript_missplicing_seqs(reference_transcript.get_context(center, 7500), variant_transcript.get_context(center, 7500), reference_transcript.donors, reference_transcript.acceptors, threshold=threshold, engine=engine)
 
-    ref_seq = 'N'*ref_start_pad + ref.seq + 'N'*ref_end_pad
-    var_seq = 'N'*var_start_pad + var.seq + 'N'*var_end_pad
 
-    if engine == 'spliceai':
-        from .spliceai_utils import sai_predict_probs, sai_models
-        ref_seq_acceptor_probs, ref_seq_donor_probs = sai_predict_probs(ref_seq, models=sai_models)
-        mut_seq_acceptor_probs, mut_seq_donor_probs = sai_predict_probs(var_seq, models=sai_models)
+    # from functools import reduce
+    # ref = transcript.pre_mrna
+    # mutations = [mutation for mutation in mutations if mutation.position in ref.indices]
+    # if len(mutations) == 0:
+    #     return {'missed_acceptors': {}, 'missed_donors': {}, 'discovered_acceptors': {}, 'discovered_donors': {}}
+    #
+    # var = reduce(lambda acc, mutation: acc + mutation, mutations, ref)
+    # center = int(np.mean([mutation.position for mutation in mutations]) // 1)
+    #
+    # total_context = context + window
+    # length = ref.seqmat.shape[-1]
+    # center_index = ref.rel_pos(center)
+    # ref_start_pad = max(0, total_context - center_index)
+    # ref_end_pad = max(0, total_context - (length - center_index))
+    #
+    # length = var.seqmat.shape[-1]
+    # center_index = var.rel_pos(center)
+    # if center_index is None:
+    #     raise IndexError("Center index must not be none... Issue with mutations... They must not be within the transcript.")
+    #
+    # var_start_pad = max(0, total_context - center_index)
+    # var_end_pad = max(0, total_context - (length - center_index))
+    #
+    # ref = ref.inspect(center, context=total_context)
+    # var = var.inspect(center, context=total_context)
+    #
+    # ref_indices = np.concatenate([np.zeros(ref_start_pad), ref.indices, np.zeros(ref_end_pad)])
+    # mut_indices = np.concatenate([np.zeros(var_start_pad),  var.indices, np.zeros(var_end_pad)])
+    #
+    # ref_indices = ref_indices[context:-context]
+    # mut_indices = mut_indices[context:-context]
+    #
+    # ref_seq = 'N'*ref_start_pad + ref.seq + 'N'*ref_end_pad
+    # var_seq = 'N'*var_start_pad + var.seq + 'N'*var_end_pad
+    #
+    # if engine == 'spliceai':
+    #     from .spliceai_utils import sai_predict_probs, sai_models
+    #     ref_seq_acceptor_probs, ref_seq_donor_probs = sai_predict_probs(ref_seq, models=sai_models)
+    #     mut_seq_acceptor_probs, mut_seq_donor_probs = sai_predict_probs(var_seq, models=sai_models)
+    #
+    # elif engine == 'pangolin':
+    #     from .pangolin_utils import pangolin_predict_probs, pang_models
+    #     ref_seq_donor_probs, ref_seq_acceptor_probs = pangolin_predict_probs(ref_seq, models=pang_models, just_ss=just_ss)
+    #     mut_seq_donor_probs, mut_seq_acceptor_probs = pangolin_predict_probs(var_seq, models=pang_models, just_ss=just_ss)
+    #
+    # else:
+    #     raise ValueError(f"{engine} not implemented")
+    #
+    # visible_donors = np.intersect1d(transcript.donors, ref_indices)
+    # visible_acceptors = np.intersect1d(transcript.acceptors, ref_indices)
+    #
+    # assert len(ref_indices) == len(ref_seq_acceptor_probs), f'Reference pos ({len(ref_indices)}) not the same as probs ({len(ref_seq_acceptor_probs)})'
+    # assert len(mut_indices) == len(mut_seq_acceptor_probs), f'Mut pos ({len(mut_indices)}) not the same as probs ({len(mut_seq_acceptor_probs)})'
+    #
+    # iap, dap = find_ss_changes({p: v for p, v in list(zip(ref_indices, ref_seq_acceptor_probs))},
+    #                            {p: v for p, v in list(zip(mut_indices, mut_seq_acceptor_probs))},
+    #                            visible_acceptors,
+    #                            threshold=threshold)
+    #
+    # assert len(ref_indices) == len(ref_seq_donor_probs), 'Reference pos not the same'
+    # assert len(mut_indices) == len(mut_seq_donor_probs), 'Mut pos not the same'
+    #
+    # idp, ddp = find_ss_changes({p: v for p, v in list(zip(ref_indices, ref_seq_donor_probs))},
+    #                            {p: v for p, v in list(zip(mut_indices, mut_seq_donor_probs))},
+    #                            visible_donors,
+    #                            threshold=threshold)
+    #
+    # ref_acceptors = {a: b for a, b in list(zip(ref_indices, ref_seq_acceptor_probs))}
+    # ref_donors = {a: b for a, b in list(zip(ref_indices, ref_seq_donor_probs))}
+    #
+    # lost_acceptors = {int(p): {'absolute': np.float64(0), 'delta': round(float(-ref_acceptors[p]), 3)} for p in
+    #                   visible_acceptors if p not in mut_indices and p not in dap}
+    # lost_donors = {int(p): {'absolute': np.float64(0), 'delta': round(float(-ref_donors[p]), 3)} for p in visible_donors
+    #                if p not in mut_indices and p not in ddp}
+    # dap.update(lost_acceptors)
+    # ddp.update(lost_donors)
+    #
+    # missplicing = {'missed_acceptors': dap, 'missed_donors': ddp, 'discovered_acceptors': iap, 'discovered_donors': idp}
+    # missplicing = {outk: {float(k): v for k, v in outv.items()} for outk, outv in missplicing.items()}
+    # temp =  {outk: {int(k) if k.is_integer() else k: v for k, v in outv.items()} for outk, outv in missplicing.items()}
+    return temp
 
-    elif engine == 'pangolin':
-        from .pangolin_utils import pangolin_predict_probs, pang_models
-        ref_seq_donor_probs, ref_seq_acceptor_probs = pangolin_predict_probs(ref_seq, models=pang_models, just_ss=just_ss)
-        mut_seq_donor_probs, mut_seq_acceptor_probs = pangolin_predict_probs(var_seq, models=pang_models, just_ss=just_ss)
 
-    else:
-        raise ValueError(f"{engine} not implemented")
+def find_transcript_missplicing_seqs(ref_seq, var_seq, donors, acceptors, threshold=0.5, engine='spliceai'):
+    if ref_seq.seq == var_seq.seq:
+        return {'missed_acceptors': {}, 'missed_donors': {}, 'discovered_acceptors': {}, 'discovered_donors': {}}
 
-    visible_donors = np.intersect1d(transcript.donors, ref_indices)
-    visible_acceptors = np.intersect1d(transcript.acceptors, ref_indices)
+    ref_seq_acceptor_probs, ref_seq_donor_probs = run_splicing_engine(ref_seq.seq, engine)
+    mut_seq_acceptor_probs, mut_seq_donor_probs = run_splicing_engine(var_seq.seq, engine)
+    ref_indices = ref_seq.indices[5000:-5000]
+    mut_indices = var_seq.indices[5000:-5000]
+    visible_donors = np.intersect1d(donors, ref_indices)
+    visible_acceptors = np.intersect1d(acceptors, ref_indices)
 
-    assert len(ref_indices) == len(ref_seq_acceptor_probs), f'Reference pos ({len(ref_indices)}) not the same as probs ({len(ref_seq_acceptor_probs)})'
-    assert len(mut_indices) == len(mut_seq_acceptor_probs), f'Mut pos ({len(mut_indices)}) not the same as probs ({len(mut_seq_acceptor_probs)})'
+    assert len(ref_indices) == len(
+        ref_seq_acceptor_probs), f'Reference pos ({len(ref_indices)}) not the same as probs ({len(ref_seq_acceptor_probs)})'
+    assert len(mut_indices) == len(
+        mut_seq_acceptor_probs), f'Mut pos ({len(mut_indices)}) not the same as probs ({len(mut_seq_acceptor_probs)})'
 
     iap, dap = find_ss_changes({p: v for p, v in list(zip(ref_indices, ref_seq_acceptor_probs))},
                                {p: v for p, v in list(zip(mut_indices, mut_seq_acceptor_probs))},
                                visible_acceptors,
-                               threshold=threshold)
+                               threshold=0.1)
 
     assert len(ref_indices) == len(ref_seq_donor_probs), 'Reference pos not the same'
     assert len(mut_indices) == len(mut_seq_donor_probs), 'Mut pos not the same'
@@ -221,113 +330,147 @@ def find_transcript_missplicing(transcript, mutations, context=5000, window=2500
     idp, ddp = find_ss_changes({p: v for p, v in list(zip(ref_indices, ref_seq_donor_probs))},
                                {p: v for p, v in list(zip(mut_indices, mut_seq_donor_probs))},
                                visible_donors,
-                               threshold=threshold)
+                               threshold=0.1)
 
     ref_acceptors = {a: b for a, b in list(zip(ref_indices, ref_seq_acceptor_probs))}
     ref_donors = {a: b for a, b in list(zip(ref_indices, ref_seq_donor_probs))}
 
     lost_acceptors = {int(p): {'absolute': np.float64(0), 'delta': round(float(-ref_acceptors[p]), 3)} for p in
                       visible_acceptors if p not in mut_indices and p not in dap}
-    lost_donors = {int(p): {'absolute': np.float64(0), 'delta': round(float(-ref_donors[p]), 3)} for p in visible_donors
+    lost_donors = {int(p): {'absolute': np.float64(0), 'delta': round(float(-ref_donors[p]), 3)} for p in
+                   visible_donors
                    if p not in mut_indices and p not in ddp}
     dap.update(lost_acceptors)
     ddp.update(lost_donors)
 
-    missplicing = {'missed_acceptors': dap, 'missed_donors': ddp, 'discovered_acceptors': iap, 'discovered_donors': idp}
+    missplicing = {'missed_acceptors': dap, 'missed_donors': ddp, 'discovered_acceptors': iap,
+                   'discovered_donors': idp}
     missplicing = {outk: {float(k): v for k, v in outv.items()} for outk, outv in missplicing.items()}
-    temp =  {outk: {int(k) if k.is_integer() else k: v for k, v in outv.items()} for outk, outv in missplicing.items()}
-    return temp
+    temp = {outk: {int(k) if k.is_integer() else k: v for k, v in outv.items()} for outk, outv in
+            missplicing.items()}
+    return Missplicing(temp, threshold=threshold)
 
 
 class Missplicing:
     def __init__(self, splicing_dict, threshold=0.5):
+        """
+        Initialize a Missplicing object.
+
+        Args:
+            splicing_dict (dict): Dictionary containing splicing events and their details.
+                                  Example:
+                                  {
+                                    "missed_acceptors": {100: {"absolute": 0.0, "delta": -0.3}, ...},
+                                    "missed_donors": { ... },
+                                    "discovered_acceptors": { ... },
+                                    "discovered_donors": { ... }
+                                  }
+            threshold (float): The threshold above which a delta is considered significant.
+        """
         self.missplicing = splicing_dict
         self.threshold = threshold
 
-    # def __repr__(self):
-    #     return f'Missplicing({self.modification.mut_id}) --> {self.missplicing}'
-
     def __str__(self):
-        return self.aberrant_splicing
+        """String representation displays the filtered splicing events passing the threshold."""
+        return str(self.significant_events)
 
     def __bool__(self):
-        if self.apply_sai_threshold_alt() is not None:
-            return True
-        return False
+        """
+        Boolean evaluation: True if any event surpasses the threshold, False otherwise.
+        """
+        return self.first_significant_event() is not None
 
     def __iter__(self):
-        vals = [0]
-        for event, details in self.missplicing.items():
-            for e, d in details.items():
-                vals.append(d['delta'])
-        return iter(vals)
-
-    # def __eq__(self, alt_splicing):
-    #     flag, _ = self.check_splicing_difference(self.missplicing, alt_splicing, self.threshold)
-    #     return not flag
+        """
+        Iterate over all delta values from all events. The first yielded value is 0 (for compatibility),
+        followed by all deltas in self.missplicing.
+        """
+        yield 0
+        for details in self.missplicing.values():
+            for d in details.values():
+                yield d['delta']
 
     @property
-    def aberrant_splicing(self):
-        return self.apply_sai_threshold(self.threshold)
+    def significant_events(self):
+        """
+        Returns a filtered version of missplicing events that meet or exceed the current threshold.
+        """
+        return self.filter_by_threshold(self.threshold)
+
+    def filter_by_threshold(self, threshold=None):
+        """
+        Filter self.missplicing to only include events where abs(delta) >= threshold.
+
+        Args:
+            threshold (float, optional): The threshold to apply. Defaults to self.threshold.
+
+        Returns:
+            dict: A new dictionary with filtered events.
+        """
+        if threshold is None:
+            threshold = self.threshold
 
-    def apply_sai_threshold(self, threshold=None):
-        splicing_dict = self.missplicing
-        if not threshold:
+        return {
+            event: {
+                pos: detail for pos, detail in details.items()
+                if abs(detail['delta']) >= threshold
+            }
+            for event, details in self.missplicing.items()
+        }
+
+    def first_significant_event(self, splicing_dict=None, threshold=None):
+        """
+        Check if there is any event surpassing a given threshold and return the dictionary if found.
+
+        Args:
+            splicing_dict (dict, optional): Dictionary to check. Defaults to self.missplicing.
+            threshold (float, optional): Threshold to apply. Defaults to self.threshold.
+
+        Returns:
+            dict or None: Returns the dictionary if a delta surpasses the threshold, otherwise None.
+        """
+        if splicing_dict is None:
+            splicing_dict = self.missplicing
+        if threshold is None:
             threshold = self.threshold
 
-        new_dict = {}
-        for event, details in self.missplicing.items():
-            in_dict = {}
-            for e, d in details.items():
-                if abs(d['delta']) >= threshold:
-                    in_dict[e] = d
-                    # return splicing_dict
-            new_dict[event] = in_dict
-        return new_dict
-
-    def apply_sai_threshold_alt(self, splicing_dict=None, threshold=None):
-        splicing_dict = self.missplicing if not splicing_dict else splicing_dict
-        threshold = self.threshold if not threshold else threshold
-        for event, details in splicing_dict.items():
-            for e, d in details.items():
-                if abs(d['delta']) >= threshold:
-                    return splicing_dict
+        # Check if any event meets the threshold
+        if any(abs(detail['delta']) >= threshold for details in splicing_dict.values() for detail in details.values()):
+            return splicing_dict
         return None
 
-    def get_max_missplicing_delta(self):
-        max_delta = 0
-        for event, details in self.missplicing.items():
-            for e, d in details.items():
-                if abs(d['delta']) > max_delta:
-                    max_delta = abs(d['delta'])
-        return max_delta
-
-
-def find_transcript_splicing(transcript, engine='spliceai'):
-    ref = transcript.pre_mrna
-    ref_start_pad = 5000
-    ref_end_pad = 5000
-
-    ref_indices = ref.indices
-    ref_seq = 'N' * ref_start_pad + ref.seq + 'N' * ref_end_pad
-    if engine == 'spliceai':
-        from .spliceai_utils import sai_predict_probs, sai_models
-        ref_seq_acceptor_probs, ref_seq_donor_probs = sai_predict_probs(ref_seq, sai_models)
-
-    elif engine == 'pangolin':
-        from .pangolin_utils import pangolin_predict_probs, pang_models
-        ref_seq_donor_probs, ref_seq_acceptor_probs = pangolin_predict_probs(ref_seq, models=pang_models)
-
-    else:
-        raise ValueError(f"{engine} not implemented")
-
-    assert len(ref_seq_donor_probs) == len(ref_indices), f'{len(ref_seq_donor_probs)}  vs. {len(ref_indices)}'
-    donor_probs = {i: p for i, p in list(zip(ref_indices, ref_seq_donor_probs))}
-    donor_probs = dict(sorted(donor_probs.items(), key=lambda item: item[1], reverse=True))
-
-    acceptor_probs = {i: p for i, p in list(zip(ref_indices, ref_seq_acceptor_probs))}
-    acceptor_probs = dict(sorted(acceptor_probs.items(), key=lambda item: item[1], reverse=True))
-    return donor_probs, acceptor_probs
+    def max_delta(self):
+        """
+        Returns the maximum absolute delta found in all events.
+
+        Returns:
+            float: The maximum absolute delta, or 0 if no events.
+        """
+        deltas = [detail['delta'] for details in self.missplicing.values() for detail in details.values()]
+        return max(deltas, key=abs, default=0.0)
+
+
+# def find_transcript_splicing(transcript, engine='spliceai'):
+#     ref_indices = transcript.indices
+#     ref_seq = 'N' * 5000 + transcript.seq + 'N' * 5000
+#     if engine == 'spliceai':
+#         from .spliceai_utils import sai_predict_probs, sai_models
+#         ref_seq_acceptor_probs, ref_seq_donor_probs = sai_predict_probs(ref_seq, sai_models)
+#
+#     elif engine == 'pangolin':
+#         from .pangolin_utils import pangolin_predict_probs, pang_models
+#         ref_seq_donor_probs, ref_seq_acceptor_probs = pangolin_predict_probs(ref_seq, models=pang_models)
+#
+#     else:
+#         raise ValueError(f"{engine} not implemented")
+#
+#     assert len(ref_seq_donor_probs) == len(ref_indices), f'{len(ref_seq_donor_probs)}  vs. {len(ref_indices)}'
+#     donor_probs = {i: p for i, p in list(zip(ref_indices, ref_seq_donor_probs))}
+#     donor_probs = dict(sorted(donor_probs.items(), key=lambda item: item[1], reverse=True))
+#
+#     acceptor_probs = {i: p for i, p in list(zip(ref_indices, ref_seq_acceptor_probs))}
+#     acceptor_probs = dict(sorted(acceptor_probs.items(), key=lambda item: item[1], reverse=True))
+#     return donor_probs, acceptor_probs
 
 
 def benchmark_splicing(gene, organism='hg38', engine='spliceai'):
@@ -337,7 +480,7 @@ def benchmark_splicing(gene, organism='hg38', engine='spliceai'):
         return None, None
 
     transcript.generate_pre_mrna()
-    predicted_donor_sites, predicted_acceptor_sites = find_transcript_splicing(transcript, engine=engine)
+    predicted_donor_sites, predicted_acceptor_sites = find_transcript_splicing(transcript.pre_mrna, engine=engine)
     num_introns = len(transcript.introns)
     predicted_donors = list(predicted_donor_sites.keys())[:num_introns]
     predicted_acceptors = list(predicted_acceptor_sites.keys())[:num_introns]
@@ -346,68 +489,58 @@ def benchmark_splicing(gene, organism='hg38', engine='spliceai'):
     return len(correct_donor_preds) / num_introns, len(correct_acceptor_preds) / num_introns, len(transcript.introns)
 
 
-def missplicing(mut_id, splicing_threshold=0.5, primary_transcript=True, organism='hg38', engine='spliceai'):
-    gene = Gene(mut_id.split(':')[0], organism=organism)
-    mutation = get_mutation(mut_id, rev=gene.rev)
-    results = {}
-
-    for tid, transcript in gene.run_transcripts():
-        # if not transcript.primary_transcript and primary_transcript:
-        #     continue
-        #
-        if mutation not in transcript:
-            continue
-
-        good_tid = tid
-
-        transcript.generate_pre_mrna()
-        results[tid] = Missplicing(find_transcript_missplicing(transcript, mutation, engine=engine),
-                                   threshold=splicing_threshold)
-
-    # if len(results) == 0:
-    #     return None
-    #
-    # if primary_transcript and good_tid in results:
-    #     return results[good_tid]
-    # else:
-    #     return None
-
-    return results
-
-
 import sqlite3
 import json
+import os
+
+# Global connection and cursor (adjust to your architecture)
+# Ideally, initialize this once in your application startup code.
+DB_PATH = os.path.join(config['splicing_db'], 'mutation_data.db')
+conn = sqlite3.connect(DB_PATH, isolation_level=None)  # autocommit mode
+cursor = conn.cursor()
+
+# Create table once at startup, not in the function
+cursor.execute('''
+CREATE TABLE IF NOT EXISTS mutations (
+    tool TEXT,
+    gene TEXT,
+    mutation_id TEXT,
+    transcript_id TEXT,
+    data TEXT,
+    PRIMARY KEY (tool, gene, mutation_id, transcript_id)
+)''')
 
 def get_or_compute_splicing(tool, gene, mutation_id, transcript_id, force_recompute=False):
-    conn = sqlite3.connect(os.path.join(tool_parameters['splicing_db'], 'mutation_data.db'))
-    cursor = conn.cursor()
-    # Create table if it doesn't exist
-    cursor.execute('''CREATE TABLE IF NOT EXISTS mutations (
-                         tool TEXT,
-                         gene TEXT,
-                         mutation_id TEXT, 
-                         transcript_id TEXT, 
-                         data TEXT,
-                         PRIMARY KEY (tool, gene, mutation_id, transcript_id)
-                      )''')
-
-    # Look up entry
+    """
+    Retrieve computed splicing data for a given mutation from a database,
+    or compute and store it if not found or if force_recompute is True.
+
+    Args:
+        tool (str): Name of the tool used for computation.
+        gene (str): Gene name or identifier.
+        mutation_id (str): A unique identifier for the mutation.
+        transcript_id (str): ID for the transcript.
+        force_recompute (bool): If True, ignore cached value and recompute.
+
+    Returns:
+        dict: The computed splicing data.
+    """
+
+    # Lookup in the database
     cursor.execute('SELECT data FROM mutations WHERE tool=? AND gene=? AND mutation_id=? AND transcript_id=?',
                    (tool, gene, mutation_id, transcript_id))
     row = cursor.fetchone()
 
-    # If entry is found and force_recompute is False, return data
+    # If found and no force recompute, return cached data
     if row and not force_recompute:
         return json.loads(row[0])
 
-    # Otherwise, compute the dictionary
-    computed_data = find_transcript_missplicing(tool, mutation_id, transcript_id)  # Replace with actual function
+    # Otherwise, compute the data
+    computed_data = find_transcript_missplicing(mutation_id, transcript_id=transcript_id, engine=tool)  # Replace with your actual function
 
-    # Store in the database
+    # Store computed data in DB
     data_json = json.dumps(computed_data)
-    cursor.execute('REPLACE INTO mutations (tool, mutation_id, transcript_id, data) VALUES (?, ?, ?, ?)',
-                   (tool, mutation_id, transcript_id, data_json))
-    conn.commit()
-    conn.close()
+    cursor.execute('REPLACE INTO mutations (tool, gene, mutation_id, transcript_id, data) VALUES (?, ?, ?, ?, ?)',
+                   (tool, gene, mutation_id, transcript_id, data_json))
 
     return computed_data

{geney-1.3.2.dist-info → geney-1.3.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: geney
-Version: 1.3.2
+Version: 1.3.4
 Summary: A Python package for gene expression modeling.
 Home-page: https://github.com/nicolaslynn/geney
 Author: Nicolas Lynn

{geney-1.3.2.dist-info → geney-1.3.4.dist-info}/RECORD

@@ -10,12 +10,12 @@ geney/graphic_utils.py,sha256=oMsBpB9YeEn96gGpKh4MmtagJffWZbk-xPrIwHvkFhA,11016
 geney/gtex_utils.py,sha256=asL2lHyU5KsbWpV096vkf1Ka7hSo_RRfZqw7p5nERmE,1919
 geney/immune_utils.py,sha256=ZRni5ttrhpYBnmNr0d0ZatIbNPYs4nmQuoUO00SpsS4,5271
 geney/mutation_utils.py,sha256=C_kv2MB_L8LlhX3W2ooXjJ3uDoJ8zX1WeDtZKoBZJkI,1547
-geney/oncosplice.py,sha256=FdvuROk2G7wwLoB5lLzYia8Smw9hHZeVs-J2MUoAwlU,22106
+geney/oncosplice.py,sha256=1xphL2LeAObwUKBXgcyyKbNO9bAryKDZesK7OpUpFfA,22336
 geney/pangolin_utils.py,sha256=i5j5vEMCWOTIa1mRP2377BAhlUFZjHBzTQBips4lA_4,2934
 geney/power_utils.py,sha256=MehZFUdkJ2EFUot709yPEDxSkXmH5XevMebX2HD768A,7330
 geney/seqmat_utils.py,sha256=wzb3PX5it5bpIFQvcxyzlxfhoJTbHHbsjg0rzh05iVs,19753
 geney/spliceai_utils.py,sha256=PFIhTK8Ihrj-cv5tgRN0UFPYEmC4uxtqXSP9bBLnZRM,3077
-geney/splicing_utils.py,sha256=4xYXy_dIbqdbVfxsEj_OCuM-MsQ24gi4fIv0vQjAYcQ,19215
+geney/splicing_utils.py,sha256=6TlSeNK8BWGnm5AqYKa_qObqHQVcgYtRY08JmfK9yZ8,26338
 geney/survival_utils.py,sha256=KnAzEviMuXh6SnVXId9PgsFLSbgkduTvYoIthxN7FPA,6886
 geney/tcga_utils.py,sha256=D_BNHm-D_K408dlcJm3hzH2c6QNFjQsKvUcOPiQRk7g,17612
 geney/tis_utils.py,sha256=2makfGfVlDFVIbxzXE85AY9jmAjcNmxyIAxjvkRA5LY,7396
@@ -24,7 +24,7 @@ geney/translation_initiation/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NM
 geney/translation_initiation/tis_utils.py,sha256=AF3siFjuQH-Rs44EV-80zHdbxRMvN4woLFSHroWIETc,4448
 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-1.3.2.dist-info/METADATA,sha256=aGPdV-x5PcONjV5ylUg8rYhW0eo4Fm2HDOE8dzldpcg,994
-geney-1.3.2.dist-info/WHEEL,sha256=fS9sRbCBHs7VFcwJLnLXN1MZRR0_TVTxvXKzOnaSFs8,110
-geney-1.3.2.dist-info/top_level.txt,sha256=O-FuNUMb5fn9dhZ-dYCgF0aZtfi1EslMstnzhc5IIVo,6
-geney-1.3.2.dist-info/RECORD,,
+geney-1.3.4.dist-info/METADATA,sha256=ONsBA4xTOrs0KaNJR9pBrwlHE06WC8YUuCFfH5vV2ag,994
+geney-1.3.4.dist-info/WHEEL,sha256=fS9sRbCBHs7VFcwJLnLXN1MZRR0_TVTxvXKzOnaSFs8,110
+geney-1.3.4.dist-info/top_level.txt,sha256=O-FuNUMb5fn9dhZ-dYCgF0aZtfi1EslMstnzhc5IIVo,6
+geney-1.3.4.dist-info/RECORD,,