def test_mutant_amino_acids_in_mm10_chr9_82927102_refGT_altTG_pT441H():
# In the Isovar repository this test is weird because the VCF only
# mentions the G>T variant but doesn't include the subsequent nucleotide
# change T>G. To avoid having to think about phasing of variants I changed
# the VCF in vaxrank to contain a GT>TG variant.
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf", data_path("b16.f10/b16.f10.Phip.vcf"),
"--bam", data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
core_logic = VaxrankCoreLogic(
reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
variants=variants,
vaccine_peptide_length=15,
padding_around_mutation=5,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True,
max_vaccine_peptides_per_variant=1)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
vaccine_peptide = vaccine_peptides[0]
mutant_protein_fragment = vaccine_peptide.mutant_protein_fragment
check_mutant_amino_acids(
variant,
mutant_protein_fragment)
def test_keep_top_k_epitopes():
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf", data_path("b16.f10/b16.f10.Phip.vcf"),
"--bam", data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
keep_k_epitopes = 3
core_logic = VaxrankCoreLogic(
reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
variants=variants,
vaccine_peptide_length=15,
padding_around_mutation=5,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True,
max_vaccine_peptides_per_variant=1,
num_mutant_epitopes_to_keep=keep_k_epitopes)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
vaccine_peptide = vaccine_peptides[0]
eq_(keep_k_epitopes, len(vaccine_peptide.mutant_epitope_predictions))
# recompute the expected score, make sure the top-k argument from ranked_vaccine_peptides()
# propagated as expected
mutant_epitope_score = sum(
p.logistic_epitope_score() for p in vaccine_peptide.mutant_epitope_predictions)
assert_almost_equal(mutant_epitope_score, vaccine_peptide.mutant_epitope_score)
def test_mutant_amino_acids_in_mm10_chrX_8125624_refC_altA_pS460I():
# there are two co-occurring variants in the RNAseq data but since
# they don't happen in the same codon then we're considering the Varcode
# annotation to be correct
# TODO: deal with phasing of variants explicitly so that both
# variant positions are considered mutated
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf", data_path("b16.f10/b16.f10.Wdr13.vcf"),
"--bam", data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
core_logic = VaxrankCoreLogic(
variants=variants,
reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
vaccine_peptide_length=15,
padding_around_mutation=5,
max_vaccine_peptides_per_variant=1,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
eq_(
1,
len(vaccine_peptides),
"Expected 1 vaccine peptide for variant '%s' but got %d" % (
variant, len(vaccine_peptides)))
vaccine_peptide = vaccine_peptides[0]
mutant_protein_fragment = vaccine_peptide.mutant_protein_fragment
check_mutant_amino_acids(variant, mutant_protein_fragment)
def test_mutant_amino_acids_in_mm10_chr9_82927102_refGT_altTG_pT441H():
# In the Isovar repository this test is weird because the VCF only
# mentions the G>T variant but doesn't include the subsequent nucleotide
# change T>G. To avoid having to think about phasing of variants I changed
# the VCF in vaxrank to contain a GT>TG variant.
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf",
data_path("b16.f10/b16.f10.Phip.vcf"),
"--bam",
data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
core_logic = VaxrankCoreLogic(reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
variants=variants,
vaccine_peptide_length=15,
padding_around_mutation=5,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True,
max_vaccine_peptides_per_variant=1)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
vaccine_peptide = vaccine_peptides[0]
mutant_protein_fragment = vaccine_peptide.mutant_protein_fragment
check_mutant_amino_acids(variant, mutant_protein_fragment)
def predict_epitopes_from_args(args):
"""
Returns an epitope collection from the given commandline arguments.
Parameters
----------
args : argparse.Namespace
Parsed commandline arguments for Topiary
"""
mhc_model = mhc_binding_predictor_from_args(args)
variants = variant_collection_from_args(args)
gene_expression_dict = rna_gene_expression_dict_from_args(args)
transcript_expression_dict = rna_transcript_expression_dict_from_args(args)
predictor = TopiaryPredictor(
mhc_model=mhc_model,
padding_around_mutation=args.padding_around_mutation,
ic50_cutoff=args.ic50_cutoff,
percentile_cutoff=args.percentile_cutoff,
min_transcript_expression=args.rna_min_transcript_expression,
min_gene_expression=args.rna_min_gene_expression,
only_novel_epitopes=args.only_novel_epitopes,
raise_on_error=not args.skip_variant_errors)
return predictor.predict_from_variants(
variants=variants,
transcript_expression_dict=transcript_expression_dict,
gene_expression_dict=gene_expression_dict)
def test_mutant_amino_acids_in_mm10_chrX_8125624_refC_altA_pS460I():
# there are two co-occurring variants in the RNAseq data but since
# they don't happen in the same codon then we're considering the Varcode
# annotation to be correct
# TODO: deal with phasing of variants explicitly so that both
# variant positions are considered mutated
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf",
data_path("b16.f10/b16.f10.Wdr13.vcf"),
"--bam",
data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
core_logic = VaxrankCoreLogic(variants=variants,
reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
vaccine_peptide_length=15,
padding_around_mutation=5,
max_vaccine_peptides_per_variant=1,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
eq_(
1, len(vaccine_peptides),
"Expected 1 vaccine peptide for variant '%s' but got %d" %
(variant, len(vaccine_peptides)))
vaccine_peptide = vaccine_peptides[0]
mutant_protein_fragment = vaccine_peptide.mutant_protein_fragment
check_mutant_amino_acids(variant, mutant_protein_fragment)
def test_keep_top_k_epitopes():
arg_parser = make_variant_sequences_arg_parser()
args = arg_parser.parse_args([
"--vcf",
data_path("b16.f10/b16.f10.Phip.vcf"),
"--bam",
data_path("b16.f10/b16.combined.sorted.bam"),
])
reads_generator = allele_reads_generator_from_args(args)
variants = variant_collection_from_args(args)
keep_k_epitopes = 3
core_logic = VaxrankCoreLogic(reads_generator=reads_generator,
mhc_predictor=random_binding_predictor,
variants=variants,
vaccine_peptide_length=15,
padding_around_mutation=5,
min_alt_rna_reads=1,
min_variant_sequence_coverage=1,
variant_sequence_assembly=True,
max_vaccine_peptides_per_variant=1,
num_mutant_epitopes_to_keep=keep_k_epitopes)
ranked_list = core_logic.ranked_vaccine_peptides()
for variant, vaccine_peptides in ranked_list:
vaccine_peptide = vaccine_peptides[0]
eq_(keep_k_epitopes, len(vaccine_peptide.mutant_epitope_predictions))
# recompute the expected score, make sure the top-k argument from ranked_vaccine_peptides()
# propagated as expected
mutant_epitope_score = sum(
p.logistic_epitope_score()
for p in vaccine_peptide.mutant_epitope_predictions)
assert_almost_equal(mutant_epitope_score,
vaccine_peptide.mutant_epitope_score)
def variant_reads_generator_from_args(args):
variants = variant_collection_from_args(args)
samfile = samfile_from_args(args)
return reads_supporting_variants(
variants=variants,
samfile=samfile,
use_duplicate_reads=args.use_duplicate_reads,
use_secondary_alignments=not args.drop_secondary_alignments,
min_mapping_quality=args.min_mapping_quality)
def variant_reads_generator_from_args(args):
variants = variant_collection_from_args(args)
samfile = samfile_from_args(args)
return reads_supporting_variants(
variants=variants,
samfile=samfile,
use_duplicate_reads=args.use_duplicate_reads,
use_secondary_alignments=not args.drop_secondary_alignments,
min_mapping_quality=args.min_mapping_quality)
def read_evidence_generator_from_args(args):
"""
Creates a generator of (Variant, ReadEvidence) pairs from parsed
arguments.
"""
variants = variant_collection_from_args(args)
samfile = alignment_file_from_args(args)
read_creator = read_collector_from_args(args)
return read_creator.read_evidence_generator(
variants=variants,
alignment_file=samfile)
def run_isovar_from_parsed_args(args):
"""
Extract parameters from parsed arguments and use them to run Isovar
"""
variants = variant_collection_from_args(args)
read_collector = read_collector_from_args(args)
alignment_file = alignment_file_from_args(args)
protein_sequence_creator = protein_sequence_creator_from_args(args)
filter_thresholds = filter_threshold_dict_from_args(args)
return run_isovar(variants=variants,
alignment_file=alignment_file,
read_collector=read_collector,
protein_sequence_creator=protein_sequence_creator,
filter_thresholds=filter_thresholds)
def main(args_list=None):
if args_list is None:
args_list = sys.argv[1:]
args = parser.parse_args(args_list)
variants = variant_collection_from_args(args)
all_effects = variants.effects()
coding_effects = all_effects.drop_silent_and_noncoding()
coding_effects_per_variant = coding_effects.top_priority_effect_per_variant()
with StringIO() as string_io:
string_io.write("%30s: %5d\n" % ("total variants", len(variants)))
string_io.write("%30s: %5d\n" % (
"# SNVs",
sum([v.is_snv for v in variants])
))
string_io.write("%30s: %5d\n" % (
"# indels",
sum([v.is_indel for v in variants])
))
string_io.write("%30s: %5d\n" % (
"coding non-synonymous variants",
len(coding_effects_per_variant)))
string_io.write("===\n\n")
string_io.write("\nCoding variants in known cancer genes:\n")
for v, e in coding_effects_per_variant.items():
if e.gene_id in cancer_driver_gene_id_set:
string_io.write("-- %s %s (%s)\n" % (e.gene_name, e.short_description, v.short_description))
string_io.write("\nCoding variants in MHC-I presentation genes:\n")
for v, e in coding_effects_per_variant.items():
if e.gene_id in class1_mhc_gene_id_set:
string_io.write("-- %s %s (%s)\n" % (e.gene_name, e.short_description, v.short_description))
string_io.write("\nCoding variants in interferon response genes:\n")
for v, e in coding_effects_per_variant.items():
if e.gene_id in interferon_response_gene_id_set:
string_io.write("-- %s %s (%s)\n" % (e.gene_name, e.short_description, v.short_description))
text = string_io.getvalue()
print(text)
if args.output_text:
with open(args.output_text, "w") as f:
f.write(text)
def run(args_list=None):
if args_list is None:
args_list = argv[1:]
args = parser.parse_args(args_list)
print("MS-MHC version %s" % __version__)
reference_genome = genome_for_reference_name(
args.genome if args.genome else "grch37")
print("Using reference genome %s" % reference_genome)
if args.vcf or args.maf or args.variant or args.json_variants:
variants = variant_collection_from_args(args)
else:
variants = []
hits = generate_protein_sequences(
genome=reference_genome,
variants=variants,
upstream_reading_frames=args.upstream_reading_frames,
downstream_reading_frames=args.downstream_reading_frames,
skip_exons=args.skip_exons,
min_peptide_length=args.min_peptide_length,
restrict_sources_to_gene_name=args.gene_name)
if args.extract_peptides:
print("Extracting %dmer-%dmer peptides from generated sequences" %
(args.min_peptide_length, args.max_peptide_length))
sequence_dict = extract_peptides(hits,
min_length=args.min_peptide_length,
max_length=args.max_peptide_length)
else:
# make sure we don't have repeated protein sequences
sequence_dict = defaultdict(list)
for sequence_obj in hits:
sequence_dict[sequence_obj.amino_acids].append(sequence_obj)
hits = collapse_peptide_sources(sequence_dict)
decoys = generate_decoys(hits,
n_decoys=len(hits) * args.num_decoys_per_hit,
random_seed=args.random_seed)
combined_sequences = hits + decoys
print("Writing %d FASTA records (%d hits, %d decoys)" %
(len(combined_sequences), len(hits), len(decoys)))
with open(args.output, "w") as f:
for seq in progressbar(combined_sequences):
seq.write_to_fasta_file(f)
print("Done.")
def ranked_variant_list_with_metadata(args):
"""
Computes all the data needed for report generation.
Parameters
----------
args : Namespace
Parsed user args from this run
Returns a dictionary containing 3 items:
- ranked variant/vaccine peptide list
- a dictionary of command-line arguments used to generate it
- patient info object
"""
if hasattr(args, 'input_json_file'):
with open(args.input_json_file) as f:
data = serializable.from_json(f.read())
# the JSON data from the previous run will have the older args saved, which may need to
# be overridden with args from this run (which all be output related)
data['args'].update(vars(args))
# if we need to truncate the variant list based on max_mutations_in_report, do that here
if len(data['variants']) > args.max_mutations_in_report:
data['variants'] = data['variants'][:args.
max_mutations_in_report]
return data
# get various things from user args
mhc_alleles = mhc_alleles_from_args(args)
logger.info("MHC alleles: %s", mhc_alleles)
variants = variant_collection_from_args(args)
logger.info("Variants: %s", variants)
# generator that for each variant gathers all RNA reads, both those
# supporting the variant and reference alleles
reads_generator = allele_reads_generator_from_args(args)
mhc_predictor = mhc_binding_predictor_from_args(args)
ranked_list, variants_count_dict = ranked_vaccine_peptides(
reads_generator=reads_generator,
mhc_predictor=mhc_predictor,
vaccine_peptide_length=args.vaccine_peptide_length,
padding_around_mutation=args.padding_around_mutation,
max_vaccine_peptides_per_variant=args.
max_vaccine_peptides_per_mutation,
min_alt_rna_reads=args.min_alt_rna_reads,
min_variant_sequence_coverage=args.min_variant_sequence_coverage,
min_epitope_score=args.min_epitope_score,
num_mutant_epitopes_to_keep=args.num_epitopes_per_peptide,
variant_sequence_assembly=args.variant_sequence_assembly)
ranked_list_for_report = ranked_list[:args.max_mutations_in_report]
patient_info = PatientInfo(
patient_id=args.output_patient_id,
vcf_paths=variants.sources,
bam_path=args.bam,
mhc_alleles=mhc_alleles,
num_somatic_variants=len(variants),
num_coding_effect_variants=variants_count_dict[
'num_coding_effect_variants'],
num_variants_with_rna_support=variants_count_dict[
'num_variants_with_rna_support'],
num_variants_with_vaccine_peptides=variants_count_dict[
'num_variants_with_vaccine_peptides'])
# return variants, patient info, and command-line args
data = {
'variants': ranked_list_for_report,
'patient_info': patient_info,
'args': vars(args),
}
logger.info('About to save args: %s', data['args'])
# save JSON data if necessary. as of time of writing, vaxrank takes ~25 min to run,
# most of which is core logic. the formatting is super fast, and it can
# be useful to save the data to be able to iterate just on the formatting
if args.output_json_file:
with open(args.output_json_file, 'w') as f:
f.write(serializable.to_json(data))
logger.info('Wrote JSON report data to %s', args.output_json_file)
return data
def ranked_vaccine_peptides_with_metadata_from_parsed_args(args):
"""
Computes all the data needed for report generation.
Parameters
----------
args : Namespace
Parsed user args from this run
Returns a dictionary containing 3 items:
- ranked variant/vaccine peptide list
- a dictionary of command-line arguments used to generate it
- patient info object
"""
if hasattr(args, 'input_json_file'):
with open(args.input_json_file) as f:
data = serializable.from_json(f.read())
# the JSON data from the previous run will have the older args saved, which may need to
# be overridden with args from this run (which all be output related)
data['args'].update(vars(args))
# if we need to truncate the variant list based on max_mutations_in_report, do that here
if len(data['variants']) > args.max_mutations_in_report:
data['variants'] = data['variants'][:args.
max_mutations_in_report]
return data
# get various things from user args
mhc_alleles = mhc_alleles_from_args(args)
logger.info("MHC alleles: %s", mhc_alleles)
variants = variant_collection_from_args(args)
logger.info("Variants: %s", variants)
vaxrank_results = run_vaxrank_from_parsed_args(args)
variants_count_dict = vaxrank_results.variant_counts()
assert len(variants) == variants_count_dict['num_total_variants'], \
"Len(variants) is %d but variants_count_dict came back with %d" % (
len(variants), variants_count_dict['num_total_variants'])
if args.output_passing_variants_csv:
variant_metadata_dicts = vaxrank_results.variant_properties(
gene_pathway_check=GenePathwayCheck())
df = pd.DataFrame(variant_metadata_dicts)
df.to_csv(args.output_passing_variants_csv, index=False)
ranked_variants_with_vaccine_peptides = vaxrank_results.ranked_vaccine_peptides
ranked_variants_with_vaccine_peptides_for_report = \
ranked_variants_with_vaccine_peptides[:args.max_mutations_in_report]
patient_info = PatientInfo(
patient_id=args.output_patient_id,
vcf_paths=variants.sources,
bam_path=args.bam,
mhc_alleles=mhc_alleles,
num_somatic_variants=variants_count_dict['num_total_variants'],
num_coding_effect_variants=variants_count_dict[
'num_coding_effect_variants'],
num_variants_with_rna_support=variants_count_dict[
'num_variants_with_rna_support'],
num_variants_with_vaccine_peptides=variants_count_dict[
'num_variants_with_vaccine_peptides'])
# return variants, patient info, and command-line args
data = {
# TODO:
# change this field to 'ranked_variants_with_vaccine_peptides'
# but figure out how to do it in a backwards compatible way
'variants': ranked_variants_with_vaccine_peptides_for_report,
'patient_info': patient_info,
'args': vars(args),
}
logger.info('About to save args: %s', data['args'])
# save JSON data if necessary. as of time of writing, vaxrank takes ~25 min to run,
# most of which is core logic. the formatting is super fast, and it can
# be useful to save the data to be able to iterate just on the formatting
if args.output_json_file:
with open(args.output_json_file, 'w') as f:
f.write(serializable.to_json(data))
logger.info('Wrote JSON report data to %s', args.output_json_file)
return data
def ranked_variant_list_with_metadata(args):
"""
Computes all the data needed for report generation.
Parameters
----------
args : Namespace
Parsed user args from this run
Returns a dictionary containing 3 items:
- ranked variant/vaccine peptide list
- a dictionary of command-line arguments used to generate it
- patient info object
"""
if hasattr(args, 'input_json_file'):
with open(args.input_json_file) as f:
data = serializable.from_json(f.read())
# the JSON data from the previous run will have the older args saved, which may need to
# be overridden with args from this run (which all be output related)
data['args'].update(vars(args))
# if we need to truncate the variant list based on max_mutations_in_report, do that here
if len(data['variants']) > args.max_mutations_in_report:
data['variants'] = data['variants'][:args.max_mutations_in_report]
return data
# get various things from user args
mhc_alleles = mhc_alleles_from_args(args)
logger.info("MHC alleles: %s", mhc_alleles)
variants = variant_collection_from_args(args)
logger.info("Variants: %s", variants)
# generator that for each variant gathers all RNA reads, both those
# supporting the variant and reference alleles
reads_generator = allele_reads_generator_from_args(args)
mhc_predictor = mhc_binding_predictor_from_args(args)
ranked_list, variants_count_dict = ranked_vaccine_peptides(
reads_generator=reads_generator,
mhc_predictor=mhc_predictor,
vaccine_peptide_length=args.vaccine_peptide_length,
padding_around_mutation=args.padding_around_mutation,
max_vaccine_peptides_per_variant=args.max_vaccine_peptides_per_mutation,
min_alt_rna_reads=args.min_alt_rna_reads,
min_variant_sequence_coverage=args.min_variant_sequence_coverage,
min_epitope_score=args.min_epitope_score,
num_mutant_epitopes_to_keep=args.num_epitopes_per_peptide,
variant_sequence_assembly=args.variant_sequence_assembly)
ranked_list_for_report = ranked_list[:args.max_mutations_in_report]
patient_info = PatientInfo(
patient_id=args.output_patient_id,
vcf_paths=variants.sources,
bam_path=args.bam,
mhc_alleles=mhc_alleles,
num_somatic_variants=len(variants),
num_coding_effect_variants=variants_count_dict['num_coding_effect_variants'],
num_variants_with_rna_support=variants_count_dict['num_variants_with_rna_support'],
num_variants_with_vaccine_peptides=variants_count_dict['num_variants_with_vaccine_peptides']
)
# return variants, patient info, and command-line args
data = {
'variants': ranked_list_for_report,
'patient_info': patient_info,
'args': vars(args),
}
logger.info('About to save args: %s', data['args'])
# save JSON data if necessary. as of time of writing, vaxrank takes ~25 min to run,
# most of which is core logic. the formatting is super fast, and it can
# be useful to save the data to be able to iterate just on the formatting
if args.output_json_file:
with open(args.output_json_file, 'w') as f:
f.write(serializable.to_json(data))
logger.info('Wrote JSON report data to %s', args.output_json_file)
return data