Proposed Analysis: Annotate SNV table of mutation frequencies

[jharenza]

What are the scientific goals of the analysis?

Annotate the SNV TSV table of mutation frequencies per cohort+cancer group+primary/relapse as will be created in #8 for conversion to JSON format.

What methods do you plan to use to accomplish the scientific goals?

Annotate the table with headings as below:
OT_SomaticTables_SNV_CNV.xlsx

Much of this can be achieved by leveraging MAF fields corresponding to the exact variant calls. For ClinVar, we may need to download a version of the database.

Update June 30th

• Instead of COSMIC frequency, annotate as COSMIC mutation census with tier from CosmicMutantExportCensus.tsv​
• Proposed Analysis: add oncoKB annotation to consensus MAF #82
• Updated analysis: add pedcbio links to SNV tables #83
• add RMTL (Y/N)

What input data are required for this analysis?

snv-consensus-plus-hotspots.maf.tsv.gz

How long do you expect is needed to complete the analysis? Will it be a multi-step analysis?

2-3 days

Who will complete the analysis (please add a GitHub handle here if relevant)?

@logstar ?

What relevant scientific literature relates to this analysis?

[logstar]

Hi @jharenza . Thank you for the analysis description.

I have a few questions about how to generate the columns in the "Example Somatic Mutation Table" sheet of OT_SomaticTables_SNV_CNV.xlsx.

• To generate Frequency in Overall Dataset (example value "4.17%"), should I use all samples, or all independent-primary-plus samples, or other set of samples in a cancer_group/cancer_group_cohort?
• To generate Mutations/Total Samples ( example value ".1/24"), should I aggregate all variants of the corresponding gene and divide by "Total Samples"? Similarly, for "Total Samples", should I use all samples, or all independent-primary-plus samples, or other set of samples in a cancer_group/cancer_group_cohort?
• To generate Identifer hg38 (See ClinVar ex) (example value "10_102599545_G_A"), should I concatenate Chromosome, Start_Position, End_Position, Reference_Allele, and Tumor_Seq_Allele2 with "_"? I briefly went over ClinVar identifier documentation and several examples, but I could not find any identifier exactly like "10_102599545_G_A".
• To generate the following columns, do we have mapping tables available? If not, I will download them from their corresponding databases.
  • Protein Identifier/Name
  • Protein Refseq ID
  • Predicted Mutation Impact Score
  • Overall COSMIC frequency
  • OncoKB cancer gene
  • OncoKB oncogene/TS gene
• To generate Hotspot, should I use the the HotSpotAllele column in the MAF file? Update (Jun 29 17:39:21 2021): found answer at doc/data-formats.md. I will use the HotSpotAllele column in the MAF file as the Hotspot columns in OT_SomaticTables_SNV_CNV.xlsx.

I will work on the frequencies first.

[jharenza]

Hi @logstar

Let's tackle cancer_group_cohort first.

• To generate Frequency in Overall Dataset (example value "4.17%"), should I use all samples, or all independent-primary-plus samples, or other set of samples in a cancer_group/cancer_group_cohort?
• To generate Mutations/Total Samples ( example value ".1/24"), should I aggregate all variants of the corresponding gene and divide by "Total Samples"? Similarly, for "Total Samples", should I use all samples, or all independent-primary-plus samples, or other set of samples in a cancer_group/cancer_group_cohort?

Ah, good question. I hadn't recalled this field, but the way we would create this is within cancer_group_cohort, identify the unique variants per patient (right now, they are at a sample level, but you could pull out BS_IDs and the mutation metadata, then merge it with the PT_IDs, drop the BS_IDs and unique them to get patient-level variant calls. This would be the data that goes into Frequency in Overall Dataset and I think instead of Mutations/Total Samples, we should make this column two columns: Total mutations and Total patients in dataset, otherwise, it looks like it would just be the fraction that corresponds to the percent.

Then, the independent-primary will be used for Frequency in primary tumors and independent-relapse will be used for Frequency in relapse tumors. That being said, I think we need another four columns for Total primary tumors mutated and Total primary tumors in dataset Total relapse tumors mutated and Total relapse tumors in dataset. Let me update the excel file, too.

• To generate Identifer hg38 (See ClinVar ex) (example value "10_102599545_G_A"), should I concatenate Chromosome, Start_Position, End_Position, Reference_Allele, and Tumor_Seq_Allele2 with "_"? I briefly went over ClinVar identifier documentation and several examples, but I could not find any identifier exactly like "10_102599545_G_A".

I think you have this almost right; there would be no end position in the above identifier.

• To generate the following columns, do we have mapping tables available? If not, I will download them from their corresponding databases.

• Protein Identifier/Name
• Protein Refseq ID
• Predicted Mutation Impact Score
• Overall COSMIC frequency
• OncoKB cancer gene
• OncoKB oncogene/TS gene

Yes, let me update the excel file- give me a few min.

• To generate Hotspot, should I use the the HotSpotAllele column in the MAF file?

yes

[logstar]

@jharenza Thank you for the detailed reply.

I will work on cancer_group_cohort first.

I agree it is more informative to have the revised columns.

If I understand correctly, Frequency in Overall Dataset = Total mutations / Total patients in dataset. The Total mutations is the number of patients that has the corresponding variant of the row.

[jharenza]

If I understand correctly, Frequency in Overall Dataset = Total mutations / Total patients in dataset. The Total mutations is the number of patients that has the corresponding variant of the row.

yes, where total mutations is total N of that specific mutation in the dataset

[logstar]

If I understand correctly, Frequency in Overall Dataset = Total mutations / Total patients in dataset. The Total mutations is the number of patients that has the corresponding variant of the row.

yes, where total mutations is total N of that specific mutation in the dataset

Got it. Thank you for the quick reply. I will work on the analysis accordingly.

[kgaonkar6]

Just wanted to make a note from the June 30th call (feel free to update/edit)

• Instead of COSMIC frequency, annotate as COSMIC mutation census with tier from CosmicMutantExportCensus.tsv​
• add RMTL (Y/N)

[logstar]

Thank you for the notes.

I wonder where I can get the CosmicMutantExportCensus.tsv​ and RMTL table for annotation. Are they going to be available in future data releases?

[kgaonkar6]

RMTL will be (soon) provided in v6
CosmicMutantExportCensus.tsv​ ( Originally from https://cancer.sanger.ac.uk/census or a reusable version of the file with the info) will also be provided in v6

[logstar]

RMTL will be (soon) provided in v6
CosmicMutantExportCensus.tsv​ ( Originally from https://cancer.sanger.ac.uk/census or a reusable version of the file with the info) will also be provided in v6

Got it. Thank you for the quick reply.

[jharenza]

I am having problems downloading the CosmicMutantExportCensus.tsv​ file - I cannot access via the Qiagen website as they suggest, so I submitted an email asking for help there. So, for now, proceed without this annotation.

[logstar]

@jharenza Is the following unfinished task included in the Gene_type annotation? The "COSMIC genes" are listed as a data source for the genelistreference.txt at https://github.com/PediatricOpenTargets/OpenPedCan-analysis/tree/dev/analyses/fusion_filtering.

Instead of COSMIC frequency, annotate as COSMIC mutation census with tier from CosmicMutantExportCensus.tsv​

If not, I could add the required annotation for the v7 annotator and snv-frequencies module updates.

[jharenza]

No, this was to use the COSMIC mutation evidence rather than the genes. I never heard back from them, so we can add it as a future ticket and enhancement if we hear back.

[logstar]

No, this was to use the COSMIC mutation evidence rather than the genes. I never heard back from them, so we can add it as a future ticket and enhancement if we hear back.

Got it. I think we could leave this ticket open as a reference.

I will also submit two tickets for adding COSMIC mutation evidence to snv-frequencies and annotator, label them with blocked, and refer to this issue.

[runjin326]

Closing with PR45 merged.