From a675e25d885f385c6ecd708f186b2ee448b97cd1 Mon Sep 17 00:00:00 2001 From: zsloan Date: Mon, 26 Jul 2021 20:47:37 +0000 Subject: Added pairscan boolean kwarg and process_rqtl_pairscan function for reading in pairscan results + renamed process_rqtl_output to process_rqtl_mapping to distinguish between that and pairscan --- gn3/api/rqtl.py | 8 +++++--- gn3/computations/rqtl.py | 17 ++++++++++++++++- 2 files changed, 21 insertions(+), 4 deletions(-) (limited to 'gn3') diff --git a/gn3/api/rqtl.py b/gn3/api/rqtl.py index ebb746c..ff76356 100644 --- a/gn3/api/rqtl.py +++ b/gn3/api/rqtl.py @@ -25,7 +25,7 @@ run the rqtl_wrapper script and return the results as JSON # Split kwargs by those with values and boolean ones that just convert to True/False kwargs = ["model", "method", "nperm", "scale", "control_marker"] - boolean_kwargs = ["addcovar", "interval", "pstrata"] + boolean_kwargs = ["addcovar", "interval", "pstrata", "pairscan"] all_kwargs = kwargs + boolean_kwargs rqtl_kwargs = {"geno": genofile, "pheno": phenofile} @@ -48,9 +48,11 @@ run the rqtl_wrapper script and return the results as JSON "output", rqtl_cmd.get('output_file'))): os.system(rqtl_cmd.get('rqtl_cmd')) - rqtl_output['results'] = process_rqtl_output(rqtl_cmd.get('output_file')) + if "pairscan" in boolean_kwargs: + rqtl_output['results'] = process_rqtl_pairscan(rqtl_cmd.get('output_file')) + else: + rqtl_output['results'] = process_rqtl_mapping(rqtl_cmd.get('output_file')) - rqtl_output['results'] = process_rqtl_output(rqtl_cmd.get('output_file')) if int(rqtl_kwargs['nperm']) > 0: rqtl_output['perm_results'], rqtl_output['suggestive'], rqtl_output['significant'] = \ process_perm_output(rqtl_cmd.get('output_file')) diff --git a/gn3/computations/rqtl.py b/gn3/computations/rqtl.py index 0433b3f..405b743 100644 --- a/gn3/computations/rqtl.py +++ b/gn3/computations/rqtl.py @@ -47,7 +47,7 @@ output filename generated from a hash of the genotype and phenotype files } -def process_rqtl_output(file_name: str) -> List: +def process_rqtl_mapping(file_name: str) -> List: """Given an output file name, read in R/qtl results and return a List of marker objects @@ -80,6 +80,21 @@ def process_rqtl_output(file_name: str) -> List: return marker_obs +def process_rqtl_pairscan(file_name: str) -> List: + """Given an output file name, read in R/qtl pair-scan results and return + a List of Lists representing the matrix of results + + """ + results = [] + with open(os.path.join(current_app.config.get("TMPDIR", "/tmp"), + "output", file_name), "r") as the_file: + for i, line in enumerate(the_file): + if i == 0: # Skip first line + continue + line_items = line.split(",") + results.append(line_items[1:]) # Append all but first item in line + + return results def process_perm_output(file_name: str): """Given base filename, read in R/qtl permutation output and calculate -- cgit v1.2.3 From 67f064606b9c8d4314701911dc4eedbaaa025cb0 Mon Sep 17 00:00:00 2001 From: zsloan Date: Thu, 29 Jul 2021 19:46:34 +0000 Subject: Fix imports to import both process_rqtl_mapping and process_rqtl_pairscan in api/rqtl.py --- gn3/api/rqtl.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'gn3') diff --git a/gn3/api/rqtl.py b/gn3/api/rqtl.py index ff76356..b0405f4 100644 --- a/gn3/api/rqtl.py +++ b/gn3/api/rqtl.py @@ -6,7 +6,7 @@ from flask import current_app from flask import jsonify from flask import request -from gn3.computations.rqtl import generate_rqtl_cmd, process_rqtl_output, process_perm_output +from gn3.computations.rqtl import generate_rqtl_cmd, process_rqtl_mapping, process_rqtl_pairscan, process_perm_output from gn3.computations.gemma import do_paths_exist rqtl = Blueprint("rqtl", __name__) -- cgit v1.2.3 From 58343f1005cb5a6aec47db22553313cf9fa17310 Mon Sep 17 00:00:00 2001 From: zsloan Date: Tue, 3 Aug 2021 20:34:06 +0000 Subject: Create pairscan_for_figure and pairscan_for_table functions that return the Dict and List respectively used for the pair scan figure and the table showing the results --- gn3/computations/rqtl.py | 87 +++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 82 insertions(+), 5 deletions(-) (limited to 'gn3') diff --git a/gn3/computations/rqtl.py b/gn3/computations/rqtl.py index 405b743..00e1db9 100644 --- a/gn3/computations/rqtl.py +++ b/gn3/computations/rqtl.py @@ -82,19 +82,96 @@ def process_rqtl_mapping(file_name: str) -> List: def process_rqtl_pairscan(file_name: str) -> List: """Given an output file name, read in R/qtl pair-scan results and return - a List of Lists representing the matrix of results + a list of both the JSON needed for the d3panels figure and a list of results + to be used when generating the results table (which will include marker names) """ - results = [] + + figure_data = pairscan_results_for_figure(file_name) + table_data = pairscan_results_for_table(file_name) + + return [figure_data, table_data] + +def pairscan_for_figure(file_name: str) -> Dict: + """Given an output file name, read in R/qtl pair-scan results and return + the JSON needed for the d3panels figure + + """ + figure_data = {} + + # Open the file with the actual results, written as a list of lists with open(os.path.join(current_app.config.get("TMPDIR", "/tmp"), "output", file_name), "r") as the_file: + lod_results = [] for i, line in enumerate(the_file): if i == 0: # Skip first line continue - line_items = line.split(",") - results.append(line_items[1:]) # Append all but first item in line + line_items = [item.rstrip('\n') for item in line.split(",")] + lod_results.append(line_items[1:]) # Append all but first item in line + figure_data['lod'] = lod_results + + # Open the map file with the list of markers/pseudomarkers and their positions + with open(os.path.join(current_app.config.get("TMPDIR", "/tmp"), + "output", "MAP_" + file_name), "r") as the_file: + chr_list = [] + pos_list = [] + for i, line in enumerate(the_file): + if i == 0: # Skip first line + continue + line_items = [item.rstrip('\n') for item in line.split(",")] + chr_list.append(line_items[1]) + pos_list.append(line_items[2]) + figure_data['chr'] = chr_list + figure_data['pos'] = pos_list + + return figure_data + + +def pairscan_for_table(file_name: str) -> List: + """Given an output file name, read in R/qtl pair-scan results and return + a list of results to be used when generating the results table (which will include marker names) + + """ + table_data = [] + + # Open the map file with the list of markers/pseudomarkers and create list of marker obs + with open(os.path.join(current_app.config.get("TMPDIR", "/tmp"), + "output", "MAP_" + file_name), "r") as the_file: + marker_list = [] + for i, line in enumerate(the_file.readlines()[1:]): + line_items = [item.rstrip('\n') for item in line.split(",")] + this_marker = { + 'name': line_items[0], + 'chr': line_items[1][1:-1], # Strip quotes from beginning and end of chr string + 'pos': line_items[2] + } + + marker_list.append(this_marker) + + # Open the file with the actual results and write the results as + # they will be displayed in the results table + with open(os.path.join(current_app.config.get("TMPDIR", "/tmp"), + "output", file_name), "r") as the_file: + for i, line in enumerate(the_file.readlines()[1:]): + marker_1 = marker_list[i] + line_items = [item.rstrip('\n') for item in line.split(",")] + for j, item in enumerate(line_items[1:]): + marker_2 = marker_list[j] + try: + lod_score = f"{float(item):.3f}" + except: + lod_score = f"{item}" + this_line = { + 'marker1': f"{marker_1['name']}", + 'pos1': f"Chr {marker_1['chr']} @ {float(marker_1['pos']):.1f} cM", + 'lod': lod_score, + 'marker2': f"{marker_2['name']}", + 'pos2': f"Chr {marker_2['chr']} @ {float(marker_2['pos']):.1f} cM" + } + table_data.append(this_line) + + return table_data - return results def process_perm_output(file_name: str): """Given base filename, read in R/qtl permutation output and calculate -- cgit v1.2.3