biomarker type mapping

[Brian]

re. biomarker_type

The biomarker_type list is legitimate as it comes directly from MM. They report a set of biomarker types per variant.

The lack of a cna seems to be purposeful.

# Note that 'CNA' is currently left out as it is relevant only in the case
# of CGI and directed in for loop below.

However, there may be a bug in the lookup logic.

I’ve checked in a new test harvester/tests/integration/test_mutation_type.py

    def test_molecular_match_CNA():
        print '"{}"={}'.format('Copy Number Variant',
                               norm_biomarker('Copy Number Variant'))
>       assert 'copy number variant' == norm_biomarker('Copy Number Variant')
E       assert 'copy number variant' == 'polymorphism'
E         - copy number variant
E         + polymorphism

biomarker_type
compared to fig D in https://www.biorxiv.org/content/biorxiv/early/2017/06/21/140475.full.pdf

• We have many biomarker types, they have 3 cna, fusion, mutation. ?
• Which is our equivalent to CNA?

features.biomarker_type.keyword: Descending	Count
mutant	8894
snp	4931
polymorphism	1404
nonsense	1160
frameshift	1040
unspecified	293
splice	213
nonsense,insertion	171
missense	146
synonymous	141
nonsense,deletion	122
synonymous,snp	114
deletion	75
intervening sequence	60
polymorphism,snp	30
nonsense,splice	22
insertion	20
gain of function	16
polymorphism,nonsense,insertion	12
loss of heterozygosity	8
polymorphism,nonsense,deletion	8
snp,nonsense,insertion	6
fusion,snp	5
snp,polymorphism,nonsense,deletion	5
indel	4
splice,deletion	4
splice,polymorphism,nonsense,insertion	4
startloss	3
5'UTR	2
gain of function,snp	2
insertion,snp	2
polymorphism,splice	2
snp,polymorphism,nonsense,insertion	2
splice,insertion	2
3'UTR	1
nonsense,snp	1
polymorphism,nonsense	1
snp,nonsense,splice	1
splice,deletion,nonsense	1

[Jeremy Goecks]

@bwalsh Can you point me to this list on MM? I believe we can simplify this considerably; we could go down to 3 types (cnas, snps, and fusions) but I'd like to see type and subtype attributes ideally.

[Brian]

@jgoecks List of all known source biomarkers with current mapping.

I've also included a suggestion we look at sequenceontology https://bioportal.bioontology.org/ontologies/SO (in use by civic) . This would allow us to harmonize on a known standard, reduce errors and reduce maintenance.

I've included a simple parent ontology. This still produces a significant (44) number of categories. Additional reduction in categories might still be accomplished by selecting a more distant ancestor.

biomarker_ontologies.xlsx

For example,

We currently harmonize the 'Insertion' reported by molecular march as a 'nonsense'. The SO match is 'insertion' and it's parent is 'sequence_alteration'

source	source_biomarkers	existing	proposed	ontology_id	match_type	parent	parent_ontology_id
civic	Insertion	nonsense	insertion	SO:0000667	exact	sequence_alteration	SO:0001059

We currently harmonize the 'Chromosome Arm' reported by molecular march as a 'splice'. The SO match is 'chromosome_arm' and it's parent is 'chromosome_part'

source	source_biomarkers	existing	proposed	ontology_id	match_type	parent	parent_ontology_id
molecularmatch	Chromosome Arm	splice	chromosome_arm	SO:0000105	exact	chromosome_part	SO:0000830

[Brian]

@jgoecks

Regarding simplifying this list.

• We can tell from the spreadsheet above that consolidating by biomarker.parent as a 'type' will not reduce the number of terms enough.
• The images below show the ontology ancestors of a few randomly selected biomarkers.
  You can browse them here
• One alternative for consolidation is to pick the Nth earliest ancestor (in this case 2). This would produce a consolidation on:

ancestor	ontology_id
region	SO:0000001
sequence_feature	SO:0000110
feature_attribute	SO:0000733
sequence_variant	SO:0001060
variant_collection	SO:0001507
functional_variant	SO:0001536
structural_variant	SO:0001537
transcript_variant	SO:0001576
variant_quality	SO:0001761
sequence_comparison	SO:0002072

sample ontology ancestors