diff options
author | AlexanderKabui | 2022-11-09 15:43:16 +0300 |
---|---|---|
committer | AlexanderKabui | 2022-11-09 15:43:16 +0300 |
commit | 3781312b3fecc2227a7a17057bc6718a64faefb4 (patch) | |
tree | 3950b395082ff7ba03b091efd96c7280adc9c90d /wqflask | |
parent | 5b587876c144c9ab512b23b5be47f9378a0bd3b2 (diff) | |
download | genenetwork2-3781312b3fecc2227a7a17057bc6718a64faefb4.tar.gz |
delete unecessary code
Diffstat (limited to 'wqflask')
-rw-r--r-- | wqflask/wqflask/correlation/show_corr_results.py | 286 |
1 files changed, 89 insertions, 197 deletions
diff --git a/wqflask/wqflask/correlation/show_corr_results.py b/wqflask/wqflask/correlation/show_corr_results.py index f279bcc3..f3082a89 100644 --- a/wqflask/wqflask/correlation/show_corr_results.py +++ b/wqflask/wqflask/correlation/show_corr_results.py @@ -74,10 +74,10 @@ def set_template_vars(start_vars, correlation_data): return correlation_data -def apply_filters(target_trait, target_dataset, **filters): +def apply_filters(trait, target_trait, target_dataset, **filters): def __p_val_filter__(p_lower, p_upper): return not (float(trait.get('corr_coefficient', 0.0)) >= p_lower and - float(trait.get('corr_coefficient', 0.0)) <= p_upper) + float(trait.get('corr_coefficient', 0.0)) <= p_upper) def __min_filter__(min_expr): if (target_dataset['type'] in ["ProbeSet", "Publish"] and target_trait['mean']): @@ -111,8 +111,6 @@ def apply_filters(target_trait, target_dataset, **filters): return True - - # check if one of the condition is not met i.e One is True return (__p_val_filter__( @@ -144,7 +142,9 @@ def get_user_filters(start_vars): get_float(start_vars, 'p_range_upper', 1.0) ) - if ["loc_chr", "min_loc_mb", "max_location_mb"] in start_vars: + if all(keys in start_vars for keys in ["loc_chr", + "min_loc_mb", + "max_location_mb"]): location_chr = get_string(start_vars, "loc_chr") min_location_mb = get_int(start_vars, "min_loc_mb") @@ -179,80 +179,86 @@ def generate_table_metadata(all_traits, dataset_metadata, dataset_obj): return (dataset_metadata | ({trait["name"]: trait for trait in metadata})) -def populate_table(dataset_metadata, target_dataset, corr_results): - def __populate_trait__(idx, target_trait, target_dataset): - results_dict = {} - results_dict['index'] = idx + 1 # - results_dict['trait_id'] = target_trait['name'] - results_dict['dataset'] = target_dataset['name'] - results_dict['hmac'] = hmac.data_hmac( - '{}:{}'.format(target_trait['name'], target_dataset['name'])) - results_dict['sample_r'] = f"{float(trait.get('corr_coefficient',0.0)):.3f}" - results_dict['num_overlap'] = trait.get('num_overlap', 0) - results_dict['sample_p'] = f"{float(trait.get('p_value',0)):.3e}" - if target_dataset['type'] == "ProbeSet": - results_dict['symbol'] = target_trait['symbol'] - results_dict['description'] = "N/A" - results_dict['location'] = target_trait['location'] - results_dict['mean'] = "N/A" - results_dict['additive'] = "N/A" - if target_trait['description']: - results_dict['description'] = target_trait['description'] - if target_trait['mean']: - results_dict['mean'] = f"{float(target_trait['mean']):.3f}" - try: - results_dict['lod_score'] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" - except: - results_dict['lod_score'] = "N/A" - results_dict['lrs_location'] = target_trait['lrs_location'] - if target_trait['additive']: - results_dict['additive'] = f"{float(target_trait['additive']):.3f}" - results_dict['lit_corr'] = "--" - results_dict['tissue_corr'] = "--" - results_dict['tissue_pvalue'] = "--" - if this_dataset['type'] == "ProbeSet": - if 'lit_corr' in trait: - results_dict['lit_corr'] = f"{float(trait['lit_corr']):.3f}" - if 'tissue_corr' in trait: - results_dict['tissue_corr'] = f"{float(trait['tissue_corr']):.3f}" - results_dict['tissue_pvalue'] = f"{float(trait['tissue_p_val']):.3e}" - elif target_dataset['type'] == "Publish": - results_dict['abbreviation_display'] = "N/A" - results_dict['description'] = "N/A" - results_dict['mean'] = "N/A" - results_dict['authors_display'] = "N/A" - results_dict['additive'] = "N/A" - results_dict['pubmed_link'] = "N/A" - results_dict['pubmed_text'] = target_trait["pubmed_text"] - - if target_trait["abbreviation"]: - results_dict = target_trait['abbreviation'] - - if target_trait["description"] == target_trait['description']: - results_dict['description'] = target_trait['description'] - - if target_trait["mean"]: - results_dict['mean'] = f"{float(target_trait['mean']):.3f}" - - if target_trait["authors"]: - authors_list = target_trait['authors'].split(',') - results_dict['authors_display'] = ", ".join( - authors_list[:6]) + ", et al." if len(authors_list) > 6 else target_trait['authors'] - - if "pubmed_id" in target_trait: - results_dict['pubmed_link'] = target_trait['pubmed_link'] - results_dict['pubmed_text'] = target_trait['pubmed_text'] - try: - results_dict["lod_score"] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" - except ValueError: - results_dict['lod_score'] = "N/A" - else: - results_dict['lrs_location'] = target_trait['lrs_location'] +def populate_table(dataset_metadata, target_dataset, this_dataset, corr_results, filters): - return results_dict + def __populate_trait__(idx, trait): - return [__populate_trait__(idx, target_trait, target_dataset) - for (idx, target_trait) in enumerate(corr_results)] + trait_name = list(trait.keys())[0] + target_trait = dataset_metadata.get(trait_name) + trait = trait[trait_name] + if not apply_filters(trait, target_trait, target_dataset, **filters): + results_dict = {} + results_dict['index'] = idx + 1 # + results_dict['trait_id'] = target_trait['name'] + results_dict['dataset'] = target_dataset['name'] + results_dict['hmac'] = hmac.data_hmac( + '{}:{}'.format(target_trait['name'], target_dataset['name'])) + results_dict['sample_r'] = f"{float(trait.get('corr_coefficient',0.0)):.3f}" + results_dict['num_overlap'] = trait.get('num_overlap', 0) + results_dict['sample_p'] = f"{float(trait.get('p_value',0)):.3e}" + if target_dataset['type'] == "ProbeSet": + results_dict['symbol'] = target_trait['symbol'] + results_dict['description'] = "N/A" + results_dict['location'] = target_trait['location'] + results_dict['mean'] = "N/A" + results_dict['additive'] = "N/A" + if target_trait['description']: + results_dict['description'] = target_trait['description'] + if target_trait['mean']: + results_dict['mean'] = f"{float(target_trait['mean']):.3f}" + try: + results_dict['lod_score'] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" + except: + results_dict['lod_score'] = "N/A" + results_dict['lrs_location'] = target_trait['lrs_location'] + if target_trait['additive']: + results_dict['additive'] = f"{float(target_trait['additive']):.3f}" + results_dict['lit_corr'] = "--" + results_dict['tissue_corr'] = "--" + results_dict['tissue_pvalue'] = "--" + if this_dataset['type'] == "ProbeSet": + if 'lit_corr' in trait: + results_dict['lit_corr'] = f"{float(trait['lit_corr']):.3f}" + if 'tissue_corr' in trait: + results_dict['tissue_corr'] = f"{float(trait['tissue_corr']):.3f}" + results_dict['tissue_pvalue'] = f"{float(trait['tissue_p_val']):.3e}" + elif target_dataset['type'] == "Publish": + results_dict['abbreviation_display'] = "N/A" + results_dict['description'] = "N/A" + results_dict['mean'] = "N/A" + results_dict['authors_display'] = "N/A" + results_dict['additive'] = "N/A" + results_dict['pubmed_link'] = "N/A" + results_dict['pubmed_text'] = target_trait["pubmed_text"] + + if target_trait["abbreviation"]: + results_dict = target_trait['abbreviation'] + + if target_trait["description"] == target_trait['description']: + results_dict['description'] = target_trait['description'] + + if target_trait["mean"]: + results_dict['mean'] = f"{float(target_trait['mean']):.3f}" + + if target_trait["authors"]: + authors_list = target_trait['authors'].split(',') + results_dict['authors_display'] = ", ".join( + authors_list[:6]) + ", et al." if len(authors_list) > 6 else target_trait['authors'] + + if "pubmed_id" in target_trait: + results_dict['pubmed_link'] = target_trait['pubmed_link'] + results_dict['pubmed_text'] = target_trait['pubmed_text'] + try: + results_dict["lod_score"] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" + except ValueError: + results_dict['lod_score'] = "N/A" + else: + results_dict['lrs_location'] = target_trait['lrs_location'] + + return results_dict + + return [__populate_trait__(idx, trait) + for (idx, trait) in enumerate(corr_results)] def correlation_json_for_table(start_vars, correlation_data, this_trait, this_dataset, target_dataset_ob): @@ -267,132 +273,18 @@ def correlation_json_for_table(start_vars, correlation_data, this_trait, this_da this_trait = correlation_data['this_trait'] this_dataset = correlation_data['this_dataset'] target_dataset = target_dataset_ob.as_dict() - - results_list = [] + corr_results = correlation_data['correlation_results'] dataset_metadata = generate_table_metadata({name for trait in corr_results for (name, _val) in trait.items()}, correlation_data["traits_metadata"], target_dataset_ob) - min_expr = get_float(start_vars, 'min_expr') - p_range_lower = get_float(start_vars, 'p_range_lower', -1.0) - p_range_upper = get_float(start_vars, 'p_range_upper', 1.0) - - if ('loc_chr' in start_vars and - 'min_loc_mb' in start_vars and - 'max_loc_mb' in start_vars): - - location_chr = get_string(start_vars, 'loc_chr') - min_location_mb = get_int(start_vars, 'min_loc_mb') - max_location_mb = get_int(start_vars, 'max_loc_mb') - else: - location_chr = min_location_mb = max_location_mb = None - - for i, trait_dict in enumerate(corr_results): - trait_name = list(trait_dict.keys())[0] - trait = trait_dict[trait_name] - - target_trait = dataset_metadata.get(trait_name) - if (float(trait.get('corr_coefficient', 0.0)) >= p_range_lower and - float(trait.get('corr_coefficient', 0.0)) <= p_range_upper): - - if (target_dataset['type'] == "ProbeSet" or target_dataset['type'] == "Publish") and bool(target_trait['mean']): - if (min_expr != None) and (float(target_trait['mean']) < min_expr): - continue - - if start_vars['location_type'] == "gene" and (target_dataset['type'] == "ProbeSet" or target_dataset['type'] == "Geno"): - if location_chr != None and (target_trait['chr'] != location_chr): - continue - if (min_location_mb != None) and (float(target_trait['mb']) < float(min_location_mb)): - continue - if (max_location_mb != None) and (float(target_trait['mb']) > float(max_location_mb)): - continue - elif target_dataset['type'] == "ProbeSet" or target_dataset['type'] == "Publish": - if location_chr != None and (target_trait['lrs_chr'] != location_chr): - continue - if (min_location_mb != None) and (float(target_trait['lrs_mb']) < float(min_location_mb)): - continue - if (max_location_mb != None) and (float(target_trait['lrs_mb']) > float(max_location_mb)): - continue - else: - continue - else: - continue - - results_dict = {} - results_dict['index'] = i + 1 - results_dict['trait_id'] = target_trait['name'] - results_dict['dataset'] = target_dataset['name'] - results_dict['hmac'] = hmac.data_hmac( - '{}:{}'.format(target_trait['name'], target_dataset['name'])) - results_dict['sample_r'] = f"{float(trait.get('corr_coefficient',0.0)):.3f}" - results_dict['num_overlap'] = trait.get('num_overlap', 0) - results_dict['sample_p'] = f"{float(trait.get('p_value',0)):.3e}" - if target_dataset['type'] == "ProbeSet": - results_dict['symbol'] = target_trait['symbol'] - results_dict['description'] = "N/A" - results_dict['location'] = target_trait['location'] - results_dict['mean'] = "N/A" - results_dict['additive'] = "N/A" - if bool(target_trait['description']): - results_dict['description'] = target_trait['description'] - if bool(target_trait['mean']): - results_dict['mean'] = f"{float(target_trait['mean']):.3f}" - try: - results_dict['lod_score'] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" - except: - results_dict['lod_score'] = "N/A" - results_dict['lrs_location'] = target_trait['lrs_location'] - if bool(target_trait['additive']): - results_dict['additive'] = f"{float(target_trait['additive']):.3f}" - results_dict['lit_corr'] = "--" - results_dict['tissue_corr'] = "--" - results_dict['tissue_pvalue'] = "--" - if this_dataset['type'] == "ProbeSet": - if 'lit_corr' in trait: - results_dict['lit_corr'] = f"{float(trait['lit_corr']):.3f}" - if 'tissue_corr' in trait: - results_dict['tissue_corr'] = f"{float(trait['tissue_corr']):.3f}" - results_dict['tissue_pvalue'] = f"{float(trait['tissue_p_val']):.3e}" - elif target_dataset['type'] == "Publish": - results_dict['abbreviation_display'] = "N/A" - results_dict['description'] = "N/A" - results_dict['mean'] = "N/A" - results_dict['authors_display'] = "N/A" - results_dict['additive'] = "N/A" - results_dict['pubmed_link'] = "N/A" - results_dict['pubmed_text'] = target_trait["pubmed_text"] - - if bool(target_trait['abbreviation']): - results_dict['abbreviation_display'] = target_trait['abbreviation'] - if bool(target_trait['description']): - results_dict['description'] = target_trait['description'] - if bool(target_trait['mean']): - results_dict['mean'] = f"{float(target_trait['mean']):.3f}" - if bool(target_trait['authors']): - authors_list = target_trait['authors'].split(',') - if len(authors_list) > 6: - results_dict['authors_display'] = ", ".join( - authors_list[:6]) + ", et al." - else: - results_dict['authors_display'] = target_trait['authors'] - if 'pubmed_id' in target_trait: - results_dict['pubmed_link'] = target_trait['pubmed_link'] - results_dict['pubmed_text'] = target_trait['pubmed_text'] - try: - results_dict['lod_score'] = f"{float(target_trait['lrs_score']) / 4.61:.1f}" - except: - results_dict['lod_score'] = "N/A" - results_dict['lrs_location'] = target_trait['lrs_location'] - if bool(target_trait['additive']): - results_dict['additive'] = f"{float(target_trait['additive']):.3f}" - else: - results_dict['location'] = target_trait['location'] - - results_list.append(results_dict) - - return json.dumps(results_list) + results = populate_table(dataset_metadata, + target_dataset, + this_dataset, corr_results, + get_user_filters(start_vars)) + return json.dumps([result for result in results if result]) def get_formatted_corr_type(corr_type, corr_method): |