def compute_all_indexes(job, options):
""" run everything (root toil job)
first all indexes are computed,
then all comparisons (follow on)
then summary (follow on of that)
"""
# do all the indexes
baseline_set = set()
for gam in options.in_gams:
baseline = baseline_path(gam, options)
if not os.path.isfile(baseline):
raise RuntimeError("baseline {} for gam {} not found".format(baseline, gam))
if baseline not in baseline_set:
job.addChildJobFn(compute_kmer_index, baseline, options, cores=options.vg_cores)
baseline_set.add(baseline)
if graph_path(gam, options) != baseline:
job.addChildJobFn(compute_kmer_index, graph_path(gam, options), options, cores=options.vg_cores)
if augmented_vg_path(gam, options) != baseline:
job.addChildJobFn(compute_kmer_index, augmented_vg_path(gam, options), options, cores=options.vg_cores)
if linear_vg_path(gam, options) != baseline:
job.addChildJobFn(compute_kmer_index, linear_vg_path(gam, options), options, cores=options.vg_cores)
# do the comparisons
job.addFollowOnJobFn(compute_all_comparisons, options, cores=1)
def dist_table(options):
""" make the jaccard distance table by scraping together all the comparison
json files
"""
# tsv header
dist_table = "#\t{}\t\t\t\t\\t\tn".format(options.baseline)
dist_table += "#graph\tgraph_dist\tlinear_dist\taugmented_dist\tsample_dist\tdelta_linear\tdelta_augmented\tdelta_sample\n"
for gam in options.in_gams:
baseline = baseline_path(gam, options)
graph_comp_path = comp_path(baseline, graph_path(gam, options), options)
graph_dist = jaccard_dist(graph_comp_path)
aug_comp_path = comp_path(baseline, augmented_vg_path(gam, options), options)
aug_graph_dist = jaccard_dist(aug_comp_path)
lin_comp_path = comp_path(baseline, linear_vg_path(gam, options), options)
lin_graph_dist = jaccard_dist(lin_comp_path)
sam_comp_path = comp_path(baseline, sample_vg_path(gam, options), options)
sam_graph_dist = jaccard_dist(sam_comp_path)
delta_lin = lin_graph_dist - graph_dist
delta_aug = aug_graph_dist - graph_dist
delta_sam = sam_graph_dist - graph_dist
dist_table += "{}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\n".format(
os.path.splitext(os.path.basename(graph_path(gam, options)))[0],
graph_dist,
lin_graph_dist,
aug_graph_dist,
sam_graph_dist,
delta_lin,
delta_aug,
delta_sam)
with open(dist_tsv_path(options), "w") as ofile:
ofile.write(dist_table)
def compute_all_comparisons(job, options):
""" run vg compare in parallel on all the graphs,
outputting a json file for each
"""
ncores = min(2, options.vg_cores)
for gam in options.in_gams:
baseline = baseline_path(gam, options)
job.addChildJobFn(compute_comparison, baseline,
graph_path(gam, options), options, cores=ncores)
job.addChildJobFn(compute_comparison, baseline,
augmented_vg_path(gam, options), options, cores=ncores)
job.addChildJobFn(compute_comparison, baseline,
linear_vg_path(gam, options), options, cores=ncores)
def acc_table(options):
""" make the accuracy table by scraping together all the comparison
json files
"""
# tsv header
acc_table = "#\t{}\t\t\t\t\t\t\t\t\t\n".format(options.baseline)
acc_table += "#graph\tgraph_prec\tgraph_rec\tgraph_f1"
acc_table += "\tlinear_prec\tlinear_rec\tlinear_f1"
acc_table += "\taugmented_prec\taugmented_rec\taugmented_f1"
acc_table += "\tsample_prec\tsample_rec\tsample_f1"
sums = defaultdict(lambda : (0.,0.,0.,0.,0.,0.,0.,0.,0.,0.,0.,0.))
counts = defaultdict(lambda : 0.)
for gam in options.in_gams:
baseline = baseline_path(gam, options)
graph_comp_path = comp_path(baseline, graph_path(gam, options), options)
graph_acc = accuracy(graph_comp_path)
aug_comp_path = comp_path(baseline, augmented_vg_path(gam, options), options)
aug_graph_acc = accuracy(aug_comp_path)
lin_comp_path = comp_path(baseline, linear_vg_path(gam, options), options)
lin_graph_acc = accuracy(lin_comp_path)
sam_comp_path = comp_path(baseline, sample_vg_path(gam, options), options)
sam_graph_acc = accuracy(sam_comp_path)
name = graph_path(gam, options)
sums[name] = (sums[name][0] + graph_acc[0],
sums[name][1] + graph_acc[1],
sums[name][2] + graph_acc[2],
sums[name][3] + lin_graph_acc[0],
sums[name][4] + lin_graph_acc[1],
sums[name][5] + lin_graph_acc[2],
sums[name][6] + aug_graph_acc[0],
sums[name][7] + aug_graph_acc[1],
sums[name][8] + aug_graph_acc[2],
sums[name][9] + sam_graph_acc[0],
sums[name][10] + sam_graph_acc[1],
sums[name][11] + sam_graph_acc[2])
counts[name] = counts[name] + 1
for name in list(set(map(lambda x : graph_path(x, options), options.in_gams))):
acc_table += "{}\t{:.4}\t{:.4}\t{:.4}\t".format(
os.path.splitext(os.path.basename(graph_path(gam, options)))[0],
float(sums[name][0]) / float(counts[name]),
float(sums[name][1]) / float(counts[name]),
float(sums[name][2]) / float(counts[name]))
acc_table += "{:.4}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\t{:.4}\t".format(
float(sums[name][3]) / float(counts[name]),
float(sums[name][4]) / float(counts[name]),
float(sums[name][5]) / float(counts[name]),
float(sums[name][6]) / float(counts[name]),
float(sums[name][7]) / float(counts[name]),
float(sums[name][8]) / float(counts[name]))
acc_table +="{:.4}\t{:.4}\t{:.4}\n".format(
float(sums[name][9]) / float(counts[name]),
float(sums[name][10]) / float(counts[name]),
float(sums[name][11]) / float(counts[name]))
with open(acc_tsv_path(options), "w") as ofile:
ofile.write(acc_table)
