Example #1
0
def run(args):
    infmat = scipy.io.loadmat(args.infmat_file)[args.infmat_name]
    infmat_index = hnio.load_index(args.infmat_index_file)
    heat, heat_params = hnio.load_heat_json(args.heat_file)
        
    if args.perm_type == "heat":
        addtl_genes = hnio.load_genes(args.permutation_genes_file) if args.permutation_genes_file else None
        deltas = get_deltas_for_heat(infmat, infmat_index, heat, addtl_genes,
                                         args.num_permutations, args.parallel)
    elif args.perm_type == "mutations":
        deltas = get_deltas_for_mutations(args, infmat, infmat_index, heat_params)
    else:
        raise ValueError("Invalid mutation permutation type: %s" % args.perm_type)
    
    #find the multiple of the median delta s.t. the size of the largest CC in the real data
    #is <= MAX_CC_SIZE
    medianDelta = np.median(deltas[MIN_CC_SIZE])
    M, gene_index = hn.induce_infmat(infmat, infmat_index, sorted(heat.keys()))
    h = hn.heat_vec(heat, gene_index)
    sim = hn.similarity_matrix(M, h)
    
    for i in range(1, 11):
        G = hn.weighted_graph(sim, gene_index, i*medianDelta)
        max_cc_size = max([len(cc) for cc in hn.connected_components(G)])
        if max_cc_size <= MAX_CC_SIZE:
            break
    
    #and recommend running HotNet with that multiple and the next 4 multiples
    recommended_deltas = [i*medianDelta for i in range(i, i+5)]

    output_file = open(args.output_file, 'w') if args.output_file else sys.stdout
    json.dump({"parameters": vars(args), "heat_parameters": heat_params,
               "recommended_deltas": recommended_deltas}, output_file, indent=4)
    if (args.output_file): output_file.close()
Example #2
0
def run(args):
    # create output directory if doesn't exist; warn if output files already exist
    if not os.path.exists(args.output_directory):
        os.makedirs(args.output_directory)
    dir_contents = os.listdir(args.output_directory)
    if JSON_OUTPUT in dir_contents or COMPONENTS_TSV in dir_contents or SIGNIFICANCE_TSV in dir_contents:
        print("WARNING: Output directory already contains HotNet results file(s), which will be "
              "overwritten. (Ctrl-c to cancel).")
    
    # load data
    infmat = scipy.io.loadmat(args.infmat_file)[args.infmat_name]  
    infmat_index = hnio.load_index(args.infmat_index_file)
    heat, heat_params = hnio.load_heat_json(args.heat_file)
  
    # compute similarity matrix and extract connected components
    M, gene_index = hn.induce_infmat(infmat, infmat_index, sorted(heat.keys()), quiet=False)
    h = hn.heat_vec(heat, gene_index)
    sim = hn.similarity_matrix(M, h)
    G = hn.weighted_graph(sim, gene_index, args.delta)
    ccs = hn.connected_components(G, args.min_cc_size)
    
    # calculate significance
    if args.permutation_type != "none":
        if args.permutation_type == "heat":
            sizes2stats = heat_permutation_significance(args, heat, infmat, infmat_index, G)
        elif args.permutation_type == "mutations":
            if heat_params["heat_fn"] != "load_mutation_heat":
                raise RuntimeError("Heat scores must be based on mutation data to perform\
                                    significance testing based on mutation data permutation.")
            sizes2stats = mutation_permutation_significance(args, infmat, infmat_index, G, heat_params)
        else:
            raise ValueError("Unrecognized permutation type %s" % (args.permutation_type))
    
    #sort ccs list such that genes within components are sorted alphanumerically, and components
    #are sorted first by length, then alphanumerically by name of the first gene in the component 
    ccs = [sorted(cc) for cc in ccs]
    ccs.sort(key=lambda comp: comp[0])
    ccs.sort(key=len, reverse=True)
    
    # write output
    output_dict = {"parameters": vars(args), "heat_parameters": heat_params,
                   "sizes": hn.component_sizes(ccs), "components": ccs}
    if args.permutation_type != "none":
        output_dict["statistics"] = sizes2stats
        hnio.write_significance_as_tsv(os.path.abspath(args.output_directory) + "/" + SIGNIFICANCE_TSV,
                                       sizes2stats)
    
    json_out = open(os.path.abspath(args.output_directory) + "/" + JSON_OUTPUT, 'w')
    json.dump(output_dict, json_out, indent=4)
    json_out.close()
    
    hnio.write_components_as_tsv(os.path.abspath(args.output_directory) + "/" + COMPONENTS_TSV, ccs)
Example #3
0
def significance_wrapper((infmat, index2gene, heat_permutation, delta, sizes)):
    M, index2gene = hn.induce_infmat(infmat, index2gene, sorted(heat_permutation.keys()))
    h = hn.heat_vec(heat_permutation, index2gene)
    sim_mat = hn.similarity_matrix(M, h)
    G = hn.weighted_graph(sim_mat, index2gene, delta)
    return num_components_min_size(G, sizes)
Example #4
0
def run(args):
    # create output directory if doesn't exist; warn if it exists and is not empty
    if not os.path.exists(args.output_directory):
        os.makedirs(args.output_directory)
    if len(os.listdir(args.output_directory)) > 0:
        print("WARNING: Output directory is not empty. Any conflicting files will be overwritten. "
              "(Ctrl-c to cancel).")
    
    infmat = scipy.io.loadmat(args.infmat_file)[INFMAT_NAME]
    infmat_index = hnio.load_index(args.infmat_index_file)
    heat = hnio.load_heat_tsv(args.heat_file)
    
    #filter out genes with heat score less than min_heat_score
    heat, addtl_genes, args.min_heat_score = hnheat.filter_heat(heat, args.min_heat_score)
    
    #find delta that maximizes # CCs of size >= MIN_SIZE for each permuted data set
    deltas = ft.get_deltas_for_heat(infmat, infmat_index, heat, addtl_genes, args.num_permutations,
                                    args.parallel)

    #find the multiple of the median delta s.t. the size of the largest CC in the real data
    #is <= MAX_CC_SIZE
    medianDelta = np.median(deltas[MIN_CC_SIZE])
    M, gene_index = hn.induce_infmat(infmat, infmat_index, sorted(heat.keys()), quiet=False)
    h = hn.heat_vec(heat, gene_index)
    sim = hn.similarity_matrix(M, h)
    
    for i in range(1, 11):
        G = hn.weighted_graph(sim, gene_index, i*medianDelta)
        max_cc_size = max([len(cc) for cc in hn.connected_components(G)])
        if max_cc_size <= MAX_CC_SIZE:
            break
    
    #and run HotNet with that multiple and the next 4 multiples
    run_deltas = [i*medianDelta for i in range(i, i+5)]
    for delta in run_deltas: 
        #create output directory
        delta_out_dir = args.output_directory + "/delta_" + str(delta)
        if not os.path.isdir(delta_out_dir):
            os.mkdir(delta_out_dir)
        
        #find connected components
        G = hn.weighted_graph(sim, gene_index, delta)
        ccs = hn.connected_components(G, args.min_cc_size)
        
        # calculate significance (using all genes with heat scores)
        print "* Performing permuted heat statistical significance..."
        heat_permutations = p.permute_heat(heat, args.num_permutations, addtl_genes, args.parallel)
        sizes = range(2, 11)
        print "\t- Using no. of components >= k (k \\in",
        print "[%s, %s]) as statistic" % (min(sizes), max(sizes))
        sizes2counts = stats.calculate_permuted_cc_counts(infmat, infmat_index, heat_permutations,
                                                          delta, sizes, args.parallel)
        real_counts = stats.num_components_min_size(G, sizes)
        size2real_counts = dict(zip(sizes, real_counts))
        sizes2stats = stats.compute_statistics(size2real_counts, sizes2counts, args.num_permutations)
    
        #sort ccs list such that genes within components are sorted alphanumerically, and components
        #are sorted first by length, then alphanumerically by name of the first gene in the component 
        ccs = [sorted(cc) for cc in ccs]
        ccs.sort(key=lambda comp: comp[0])
        ccs.sort(key=len, reverse=True)
    
        # write output
        output_dict = {"parameters": vars(args), "sizes": hn.component_sizes(ccs),
                       "components": ccs, "statistics": sizes2stats}
        hnio.write_significance_as_tsv(os.path.abspath(delta_out_dir) + "/" + SIGNIFICANCE_TSV,
                                       sizes2stats)
        
        json_out = open(os.path.abspath(delta_out_dir) + "/" + JSON_OUTPUT, 'w')
        json.dump(output_dict, json_out, indent=4)
        json_out.close()
        
        hnio.write_components_as_tsv(os.path.abspath(delta_out_dir) + "/" + COMPONENTS_TSV, ccs)
Example #5
0
def run(args):
    # create output directory if doesn't exist; warn if it exists and is not empty
    if not os.path.exists(args.output_directory):
        os.makedirs(args.output_directory)
    if len(os.listdir(args.output_directory)) > 0:
        print("WARNING: Output directory is not empty. Any conflicting files will be overwritten. "
              "(Ctrl-c to cancel).")
    
    infmat = scipy.io.loadmat(args.infmat_file)[INFMAT_NAME]
    infmat_index = hnio.load_index(args.infmat_index_file)
    heat = hnio.load_heat_tsv(args.heat_file)
    
    # filter out genes with heat score less than min_heat_score
    heat, addtl_genes, args.min_heat_score = hnheat.filter_heat(heat, args.min_heat_score)
    
    # find delta that maximizes # CCs of size >= MIN_SIZE for each permuted data set
    deltas = ft.get_deltas_for_heat(infmat, infmat_index, heat, addtl_genes, args.num_permutations,
                                    args.parallel)

    #find the multiple of the median delta s.t. the size of the largest CC in the real data
    #is <= MAX_CC_SIZE
    medianDelta = np.median(deltas[MIN_CC_SIZE])
    M, gene_index = hn.induce_infmat(infmat, infmat_index, sorted(heat.keys()), quiet=False)
    h = hn.heat_vec(heat, gene_index)
    sim = hn.similarity_matrix(M, h)
    
    for i in range(1, 11):
        G = hn.weighted_graph(sim, gene_index, i*medianDelta)
        max_cc_size = max([len(cc) for cc in hn.connected_components(G)])
        if max_cc_size <= MAX_CC_SIZE:
            break
    
    # load interaction network edges and determine location of static HTML files for visualization
    edges = hnio.load_ppi_edges(args.edge_file) if args.edge_file else None
    index_file = '%s/viz_files/%s' % (os.path.realpath(__file__).rsplit('/', 1)[0], VIZ_INDEX)
    subnetworks_file = '%s/viz_files/%s' % (os.path.realpath(__file__).rsplit('/', 1)[0], VIZ_SUBNETWORKS)
    gene2index = dict([(gene, index) for index, gene in infmat_index.iteritems()])

    #and run HotNet with that multiple and the next 4 multiples
    run_deltas = [i*medianDelta for i in range(i, i+5)]
    for delta in run_deltas: 
        # create output directory
        delta_out_dir = args.output_directory + "/delta_" + str(delta)
        if not os.path.isdir(delta_out_dir):
            os.mkdir(delta_out_dir)
        
        # find connected components
        G = hn.weighted_graph(sim, gene_index, delta)
        ccs = hn.connected_components(G, args.min_cc_size)
        
        # calculate significance (using all genes with heat scores)
        print "* Performing permuted heat statistical significance..."
        heat_permutations = p.permute_heat(heat, args.num_permutations, addtl_genes, args.parallel)
        sizes = range(2, 11)
        print "\t- Using no. of components >= k (k \\in",
        print "[%s, %s]) as statistic" % (min(sizes), max(sizes))
        sizes2counts = stats.calculate_permuted_cc_counts(infmat, infmat_index, heat_permutations,
                                                          delta, sizes, args.parallel)
        real_counts = stats.num_components_min_size(G, sizes)
        size2real_counts = dict(zip(sizes, real_counts))
        sizes2stats = stats.compute_statistics(size2real_counts, sizes2counts, args.num_permutations)
    
        # sort ccs list such that genes within components are sorted alphanumerically, and components
        # are sorted first by length, then alphanumerically by name of the first gene in the component 
        ccs = [sorted(cc) for cc in ccs]
        ccs.sort(key=lambda comp: comp[0])
        ccs.sort(key=len, reverse=True)
    
        # write output
        heat_dict = {"heat": heat, "parameters": {"heat_file": args.heat_file}}
        heat_out = open(os.path.abspath(delta_out_dir) + "/" + HEAT_JSON, 'w')
        json.dump(heat_dict, heat_out, indent=4)
        heat_out.close()
        
        args.heat_file = os.path.abspath(delta_out_dir) + "/" + HEAT_JSON
        args.delta = delta
        output_dict = {"parameters": vars(args), "sizes": hn.component_sizes(ccs),
                       "components": ccs, "statistics": sizes2stats}
        hnio.write_significance_as_tsv(os.path.abspath(delta_out_dir) + "/" + SIGNIFICANCE_TSV,
                                       sizes2stats)
        
        json_out = open(os.path.abspath(delta_out_dir) + "/" + JSON_OUTPUT, 'w')
        json.dump(output_dict, json_out, indent=4)
        json_out.close()
        
        hnio.write_components_as_tsv(os.path.abspath(delta_out_dir) + "/" + COMPONENTS_TSV, ccs)

        # write visualization output if edge file given
        if args.edge_file:
            viz_data = {"delta": delta, 'subnetworks': list()}
            for cc in ccs:
                viz_data['subnetworks'].append(viz.get_component_json(cc, heat, edges, gene2index,
                                                                      args.network_name))
                
            delta_viz_dir = '%s/viz/delta%s' % (args.output_directory, delta)
            if not os.path.isdir(delta_viz_dir):
                os.makedirs(delta_viz_dir)
            viz_out = open('%s/subnetworks.json' % delta_viz_dir, 'w')
            json.dump(viz_data, viz_out, indent=4)
            viz_out.close()
   
            shutil.copy(subnetworks_file, delta_viz_dir)
    
    if args.edge_file:
        viz.write_index_file(index_file, '%s/viz/%s' % (args.output_directory, VIZ_INDEX), run_deltas)