Update mutation axis datasets and documentation

README.md

@@ -5,18 +5,42 @@ configs:
   data_files:
   - split: train
     path: mondo_ontology/train-00000-of-00001.parquet
-- config_name: mutated_genes
+- config_name: tcga_mutations
   data_files:
   - split: train
-    path: mutations/mutated_genes.parquet
-- config_name: assayed_samples
+    path: mutations/tcga_mutations.parquet
+- config_name: tcga_clinical
   data_files:
   - split: train
-    path: mutations/assayed_samples.parquet
-- config_name: assayed_genes
+    path: mutations/tcga_clinical.parquet
+- config_name: depmap_mutations
   data_files:
   - split: train
-    path: mutations/assayed_genes.parquet
+    path: mutations/depmap_mutations.parquet
+- config_name: depmap_subtypes
+  data_files:
+  - split: train
+    path: mutations/depmap_subtypes.parquet
+- config_name: depmap_copy_number
+  data_files:
+  - split: train
+    path: mutations/depmap_copy_number.parquet
+- config_name: depmap_features
+  data_files:
+  - split: train
+    path: mutations/depmap_features.parquet
+- config_name: oncotree
+  data_files:
+  - split: train
+    path: mutations/oncotree.parquet
+- config_name: cell_lines
+  data_files:
+  - split: train
+    path: mutations/cell_lines.parquet
+- config_name: metaprism_clinical
+  data_files:
+  - split: train
+    path: mutations/metaprism_clinical.parquet
 task_categories:
 - other
 tags:
@@ -38,7 +62,7 @@ The collection brings together patient samples, patient-derived xenografts (PDXs
 | Component | Contents | Format |
 |---|---|---|
 | Expression | Five datasets with TPM values, sample metadata and MONDO labels | AnnData / H5AD |
-| Mutations | Gene-level mutation presence and sample/gene reference tables for TCGA, META-PRISM and DepMap | Parquet |
+| Mutations | Nine source tables for variant calls, clinical annotations, copy number and ontology evidence | Parquet |
 | Disease hierarchy | 15,481 MONDO terms and 25,555 parent relations | Parquet |
 
 ## Expression datasets
@@ -77,31 +101,29 @@ Each H5AD contains a sample-by-gene matrix. All five datasets use the same order
 
 ## Mutation data
 
-Three complementary tables connect mutation data directly to the expression samples. The `dataset` column contains `tcga`, `metaprism` or `depmap`.
+**OncoTree version:** all 897 exported codes and levels match the official release `oncotree_2025_10_03` (3 October 2025).
 
-| File | One row represents | Columns | Rows |
-|---|---|---|---:|
-| [mutated_genes.parquet](mutations/mutated_genes.parquet) | A sample–gene pair with a qualifying mutation | `dataset`, `sample_id`, `gene_id_ensembl`, `gene_symbol` | 345,620 |
-| [assayed_samples.parquet](mutations/assayed_samples.parquet) | An expression sample represented in the mutation matrices | `dataset`, `sample_id` | 10,696 |
-| [assayed_genes.parquet](mutations/assayed_genes.parquet) | A gene represented in a dataset's mutation matrix | `dataset`, `gene_id_ensembl` | 37,298 |
+Nine tables support the mutation axis. Values and identifiers were preserved from the processed sources.
 
-| Dataset | Samples with mutation data | Mutated sample–gene pairs | Genes in the mutation matrix |
-|---|---:|---:|---:|
-| TCGA | 8,877 | 244,203 | 18,566 |
-| META-PRISM | 354 | 567 | 124 |
-| DepMap | 1,465 | 100,850 | 18,608 |
+| File | Contents | Retained columns | Rows |
+|---|---|---|---:|
+| [tcga_mutations.parquet](mutations/tcga_mutations.parquet) | TCGA somatic variant calls | sample_id, gene_id_ensembl, variant_classification, aa_change | 2,845,982 |
+| [tcga_clinical.parquet](mutations/tcga_clinical.parquet) | TCGA OncoTree and MSI annotations | sample_id, cancer_type_oncotree, msi_score_mantis, msi_score_sensor | 10,953 |
+| [depmap_mutations.parquet](mutations/depmap_mutations.parquet) | DepMap 25Q3 somatic variant calls | Consensus_model_id, is_default_entry_for_model, gene_id_ensembl, protein_change, variant_info, ref, alt, gt, ps | 750,004 |
+| [depmap_subtypes.parquet](mutations/depmap_subtypes.parquet) | DepMap 25Q3 OncoTree and MSI subtype tags | Consensus_model_id, Subtype | 5,828 |
+| [depmap_copy_number.parquet](mutations/depmap_copy_number.parquet) | DepMap 25Q3 copy-number vectors | Consensus_model_id, omics_features | 1,095 |
+| [depmap_features.parquet](mutations/depmap_features.parquet) | DepMap 25Q3 copy-number feature identifiers | copy_number_features | 1 |
+| [oncotree.parquet](mutations/oncotree.parquet) | OncoTree cancer codes and levels | oncotree_code, oncotree_level | 897 |
+| [cell_lines.parquet](mutations/cell_lines.parquet) | Cellosaurus MSI and mutation annotations for DepMap lines | cvcl_id, msi_status, mutations | 1,507 |
+| [metaprism_clinical.parquet](mutations/metaprism_clinical.parquet) | META-PRISM patient, sample and OncoTree annotations | patient_id, sample_id, cancer_type_oncotree | 1,031 |
 
-**Mutation definition** 
-Variants with `vep_impact` equal to `HIGH` or `MODERATE` were retained. Multiple qualifying variants in a gene were collapsed to binary presence. These tables describe mutation presence, rather than individual variants or allele frequencies.
+**Copy number.** Each omics_features vector is indexed by the ordered copy_number_features list in depmap_features.parquet. Read the files together.
 
-**Sample matching** 
-Join by `dataset` and `sample_id`; `sample_id` matches the observation identifier in the corresponding H5AD. TCGA patient-level calls and META-PRISM subject-level calls were assigned to their expression samples. DepMap calls were matched to model identifiers. DepMap mutation data use **25Q3**, while expression data use **24Q4**.
+**Sample matching.** TCGA uses sample_id, DepMap uses Consensus_model_id, Cellosaurus uses cvcl_id, and META-PRISM supplies patient_id and sample_id. Apply the benchmark's cohort-specific matching rules. 
 
-**Missing data** 
-A sample present in `assayed_samples` but absent from `mutated_genes` has no qualifying mutation in the represented gene set. A sample absent from `assayed_samples` has no mutation data in this export.
+**Missing evidence.** Missing calls or annotations alone do not imply a negative biomarker label. Coverage and eligibility are determined by the mutation-axis code.
 
-**Comparison panel** 
-The intersection of the three mutation gene sets contains **124 genes**. The `assayed_genes` table records columns present in the input mutation matrices. 
+**External META-PRISM mutations.** The full MAF remains on the authors' public server. The benchmark downloads it and checks the SHA-256 recorded in mutation_sources.yaml; it is not duplicated here.
 
 ## MONDO disease hierarchy
 
@@ -128,10 +150,12 @@ The non-human animal disease branch (`MONDO:0005583` and 62 exclusive descendant
 | TCGA expression | NCI/NHGRI; [Genomic Data Commons](https://portal.gdc.cancer.gov/) |
 | DepMap expression | Broad Institute; [DepMap 24Q4 Public](https://doi.org/10.25452/figshare.plus.27993248) |
 | GSE317901 expression | Queen's University Belfast and collaborators; [GEO GSE317901](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE317901) |
-| TCGA mutations | [cBioPortal Datahub](https://github.com/cBioPortal/datahub/tree/master/public), `*tcga_pan_can_atlas_2018*/data_mutations.txt` |
-| META-PRISM mutations | [Publication supplementary material](https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/#sec63), Table S6 |
-| DepMap mutations | Broad Institute; [DepMap Public 25Q3](https://depmap.org/portal/download/all/), `OmicsSomaticMutations.csv` |
+| TCGA mutations and clinical annotations | [cBioPortal Datahub](https://github.com/cBioPortal/datahub/tree/master/public), `*tcga_pan_can_atlas_2018*/data_mutations.txt` |
+| META-PRISM clinical annotations | [Publication supplementary material](https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/#sec63), Table S1 |
+| DepMap mutation-axis inputs | Broad Institute; [DepMap Public 25Q3](https://depmap.org/portal/download/all/); somatic calls, subtype tags, copy-number measurements and feature identifiers |
 | MONDO | Monarch Initiative; [Mondo Disease Ontology](https://mondo.monarchinitiative.org/); [official OBO reference](https://purl.obolibrary.org/obo/mondo.obo) |
+| Cellosaurus | SIB Swiss Institute of Bioinformatics; [Cellosaurus](https://www.cellosaurus.org/); [doi:10.7171/jbt.18-2902-002](https://doi.org/10.7171/jbt.18-2902-002). |
+| OncoTree | Memorial Sloan Kettering Cancer Center; [OncoTree](https://oncotree.mskcc.org/), release oncotree_2025_10_03 |
 
 ### Licence and reuse
 
@@ -141,3 +165,15 @@ The collection does not assign a single licence to all components.
 The PDXE deposit, DepMap 24Q4 expression release and MONDO ontology identify **CC BY 4.0**. META-PRISM processed expression tables were made publicly available by the authors. TCGA expression was obtained through GDC open-access resources. GEO data are subject to [NCBI's data-use notice](https://www.ncbi.nlm.nih.gov/geo/info/disclaimer.html).
 
 
+
+| Mutation-axis files | Licence and reuse |
+|---|---|
+| tcga_mutations.parquet, tcga_clinical.parquet | Adapted tables offered under [ODbL 1.0](https://opendatacommons.org/licenses/odbl/1-0/) with [cBioPortal attribution](https://github.com/cBioPortal/datahub#license). The datahub also includes a TCGA-specific GDAC notice; applicable study terms require release review. |
+| Four depmap_* tables | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) |
+| cell_lines.parquet | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/), as stated by [Cellosaurus](https://www.cellosaurus.org/description.html). Rows and columns were selected for this export. |
+| oncotree.parquet | [CC BY 4.0](https://github.com/cBioPortal/oncotree/blob/master/LICENSE.md). |
+| metaprism_clinical.parquet | The [publication](https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/) states CC BY-NC-ND 4.0. |
+
+Contains information from the cBioPortal datahub, which is made available under the Open Database License (ODbL) 1.0.
+
+The ODbL designation above applies to the two identified TCGA tables; other components retain their source-specific terms.

dataset_register.json

@@ -43,8 +43,8 @@
     "license": "The processed expression tables used in this dataset were made publicly available by the META-PRISM authors.",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. 932 samples are assessed, 932 retained and 0 excluded for missing/invalid MONDO IDs. META-PRISM mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-MUTATED-GENES, DS-2026-MAPPING-ASSAYED-SAMPLES and DS-2026-MAPPING-ASSAYED-GENES.",
-    "legal_basis_and_rights":"n/a",
+    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. 932 samples are assessed, 932 retained and 0 excluded for missing/invalid MONDO IDs. META-PRISM mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-METAPRISM-CLINICAL; the full MAF is obtained separately from the authors’ public server.",
+    "legal_basis_and_rights": "n/a",
     "quality_criteria": "Verified on the v1 export: all 932 retained rows have a syntactically valid MONDO ID; all 19260 genes and their order are unchanged; sample identifiers are unique and retained in source order; expression is finite and nonnegative; the output contains only the benchmark metadata allowlist and no auxiliary AnnData slots. Gene audit: 79 artificial columns, all zero; panel membership and ordering match the inspected reference exactly.",
     "processing_steps": [
       "Authors’ data: The source data used were META-PRISM gene TPM table and supplementary clinical tables, including Primary_Tumor_Site, Histology, Biopsy_Site. MONDO and tissue-ontology columns are not author-supplied fields.",
@@ -73,7 +73,7 @@
     "license": "GDC open-access",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. From our annotated matrix, 11504 samples are assessed, 11425 retained and 79 excluded for missing/invalid MONDO IDs. TCGA mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-MUTATED-GENES, DS-2026-MAPPING-ASSAYED-SAMPLES and DS-2026-MAPPING-ASSAYED-GENES.",
+    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. From our annotated matrix, 11504 samples are assessed, 11425 retained and 79 excluded for missing/invalid MONDO IDs. TCGA mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-TCGA-MUTATIONS and DS-2026-MAPPING-TCGA-CLINICAL.",
     "legal_basis_and_rights": "n/a",
     "quality_criteria": "Verified on the v1 export: all 11425 retained rows have a syntactically valid MONDO ID; all 19260 genes and their order are unchanged; sample identifiers are unique and retained in source order; expression is finite and nonnegative; the output contains only the benchmark metadata allowlist and no auxiliary AnnData slots. Gene audit: 29 artificial columns, all zero; panel membership and ordering match the inspected reference exactly.",
     "processing_steps": [
@@ -103,9 +103,9 @@
     "license": "DepMap 24Q4 Public, DOI 10.25452/figshare.plus.27993248, is listed as CC BY 4.0 on its release deposit.",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. From our annotated matrix, 1673 samples are assessed, 1508 retained and 165 excluded for missing/invalid MONDO IDs; all 19260 genes are retained. DepMap mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-MUTATED-GENES, DS-2026-MAPPING-ASSAYED-SAMPLES and DS-2026-MAPPING-ASSAYED-GENES.",
+    "provenance": "First benchmark release v1 (2026-09-11). Experimental expression and study metadata originate from the authors/resources named in source. Our team harmonizes metadata and adds MONDO disease annotations as described in processing_steps. From our annotated matrix, 1673 samples are assessed, 1508 retained and 165 excluded for missing/invalid MONDO IDs; all 19260 genes are retained. DepMap mutation data were also used in the benchmark and are documented separately under DS-2026-MAPPING-DEPMAP-MUTATIONS, DS-2026-MAPPING-DEPMAP-SUBTYPES, DS-2026-MAPPING-DEPMAP-COPY-NUMBER and DS-2026-MAPPING-DEPMAP-FEATURES.",
     "legal_basis_and_rights": "n/a",
-    "quality_criteria": "Verified on the v1 export: all 1508 retained rows have a syntactically valid MONDO ID; all 19260 genes and their order are unchanged; sample identifiers are unique and retained in source order; expression is finite and nonnegative; the output contains only the benchmark metadata allowlist and no auxiliary AnnData slots. The retained DepMap set contains 6 samples with two syntactically valid MONDO IDs absent from the pinned benchmark graph (MONDO:0019210 and MONDO:0850271). Gene audit: 285 artificial columns, all zero.",
+    "quality_criteria": "Verified on the v1 export: all 1508 retained rows have a syntactically valid MONDO ID; all 19260 genes and their order are unchanged; sample identifiers are unique and retained in source order; expression is finite and nonnegative; the output contains only the benchmark metadata allowlist and no auxiliary AnnData slots. The retained DepMap set contains 6 samples with two syntactically valid MONDO IDs absent from the benchmark graph (MONDO:0019210 and MONDO:0850271). Gene audit: 285 artificial columns, all zero.",
     "processing_steps": [
       "Authors’ data: The source data used were DepMap 24Q4 OmicsExpressionProteinCodingGenesTPMLogp1.csv and release sample/profile metadata. MONDO and tissue-ontology columns are not author-supplied fields.",
       "Expression processing: Expression values were converted from log2(TPM+1) to TPM. Expression values were stored in CSR format with Ensembl gene identifiers.",
@@ -150,109 +150,274 @@
     "last_reviewed": "2026-09-11"
   },
   {
-    "dataset_id": "DS-2026-MAPPING-MUTATED-GENES",
-    "dataset_name": "Non-silent mutated genes by expression sample",
-    "version": "v1 / 2026-09-15;",
+    "dataset_id": "DS-2026-MAPPING-MONDO",
+    "dataset_name": "MONDO disease ontology",
+    "version": "v1 / 2026-09-14",
     "dataset_type": "benchmark",
     "related_ai_system": [
       "Ingenix Research Cluster"
     ],
-    "intended_purpose": "Represent mutation presence for sample matching and Jaccard evaluation over the shared gene panel.",
-    "source": "TCGA Pan-Cancer Atlas 2018 mutation files distributed through cBioPortal Datahub (https://github.com/cBioPortal/datahub/tree/master/public); Gustave Roussy META-PRISM publication supplementary Table S6 (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/#sec63); Broad Institute DepMap Public 25Q3, OmicsSomaticMutations.csv (https://depmap.org/portal/download/all/).",
-    "source_and_acquisition": "TCGA: The preparation pipeline downloaded cBioPortal Datahub study folders with Git LFS and read public/*tcga_pan_can_atlas_2018*/data_mutations.txt. Cohort mutation tables were concatenated and their variant annotations were standardized. META-PRISM: Mutation calls were read from Table S6 of the publication supplement. DepMap: Raw somatic mutation calls were read from OmicsSomaticMutations.csv for release 25Q3; download URLs were resolved through the provider endpoint https://depmap.org/portal/api/download/files. For the benchmark, variants with vep_impact HIGH or MODERATE were collapsed to gene-level mutation presence. TCGA patient-level calls and META-PRISM subject-level calls were linked to the corresponding expression samples, while DepMap models were matched by model ID. The derived exports were aligned to the expression sample identifiers.",
-    "license": "TCGAl: GDC open-access. META-PRISM: (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/). DepMap: CC by 4.0",
+    "intended_purpose": "Provide disease identifiers, names and ancestor relationships for ontology-based comparison of disease labels in the mapping benchmark.",
+    "source": "Mondo Disease Ontology (MONDO), Monarch Initiative. Official ontology reference: https://purl.obolibrary.org/obo/mondo.obo.",
+    "source_and_acquisition": "Only the MONDO component was retained in the benchmark export. The official MONDO ontology is available in OWL, OBO and JSON formats.",
+    "license": "Creative Commons Attribution 4.0 International (CC BY 4.0). MONDO licence: https://github.com/monarch-initiative/mondo/blob/master/LICENSE.",
+    "ai_augmentation": null,
+    "provenance": "The export contains 15,481 MONDO nodes, 25,555 parent relations and ten hierarchy columns. The non-human animal disease branch was excluded.",
+    "legal_basis_and_rights": "The export contains ontology identifiers, disease names and graph relationships, without patient or sample records. Reuse of MONDO ontology content is subject to CC BY 4.0",
+    "quality_criteria": "The recorded export audit confirms an acyclic graph with one root, MONDO:0000001, and preserved reachability among retained MONDO nodes. No dangling source parent references or conflicting nonempty source MONDO mappings were recorded.",
+    "processing_steps": [
+      "Source preparation: MONDO terms and relationships were prepared for the benchmark.",
+      "Source validation: The source graph was checked for cycles and dangling parent references. Nonempty MONDO identifier mappings were checked for conflicts.",
+      "Hierarchy processing: Direct parent relations between retained MONDO nodes were preserved. Paths through non-MONDO nodes were projected onto the nearest MONDO ancestors. Nine derived relations and their source paths were recorded.",
+      "Branch filtering: MONDO:0005583 and its 62 exclusive descendants were excluded. No benchmark sample carried an excluded label.",
+      "Hierarchy annotation: Parent relationships, ancestor closures, names, counts and levels were recomputed. The mondo_level field was defined as the minimum distance from a root. The original scalar parent was retained when it remained a direct parent. Otherwise, the first sorted parent was selected. The direct_ancestors_ids field was retained as the authoritative multi-parent representation.",
+      "Export schema: Ten columns were retained, covering MONDO identifiers, names, levels, parents and ancestors."
+    ],
+    "bias_assessment": "Coverage and hierarchy reflect the source ontology content and the selected MONDO subgraph.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-09-16"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-TCGA-MUTATIONS",
+    "dataset_name": "TCGA somatic variant calls",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "TCGA Pan-Cancer Atlas 2018, cBioPortal datahub; https://github.com/cBioPortal/datahub/tree/master/public",
+    "source_and_acquisition": "Mutation data_mutations.txt files were filtered upstream to FILTER == PASS and Annotation_Status == SUCCESS, according to the migration specification.",
+    "license": "ODbL 1.0 with cBioPortal attribution. The datahub also includes a TCGA-specific GDAC notice; applicable study terms require release approval. https://github.com/cBioPortal/datahub#license",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "One row per dataset, expression sample and Ensembl gene with at least one qualifying mutation. The supplied reference export contains 345620 rows: tcga 244203, depmap 100850, metaprism 567.",
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
     "legal_basis_and_rights": "n/a",
-    "quality_criteria": "Verified column names and string types; no nulls, empty key values or duplicate keys. All assayed samples exist in the corresponding expression H5ADs.",
+    "quality_criteria": "2,845,982 rows and exactly 4 columns: sample_id, gene_id_ensembl, variant_classification, aa_change. Row count matches the specification. Parquet round-trip equality was checked against the selected source values.",
     "processing_steps": [
-      "Mutation filtering: The documented non-silent definition was applied by selecting variants with HIGH or MODERATE impact. Multiple qualifying variants within a gene were collapsed to binary mutation presence.",
-      "Sample matching: Patient-level TCGA calls and subject-level META-PRISM calls were associated with their expression samples. DepMap models were matched to expression sample identifiers.",
-      "Mutation export: Positive sample–gene entries were exported. Gene symbols were attached where available from expression metadata."
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only sample_id, gene_id_ensembl, variant_classification, aa_change were retained.",
+      "Row and value selection: Two source shards were concatenated in order. All rows and genes were retained.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
     ],
-    "bias_assessment": "TCGA, META-PRISM and DepMap differ in biological models and mutation ascertainment.",
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
     "data_owner": "Alicja Augustyniak",
-    "last_reviewed": "2026-09-15"
+    "last_reviewed": "2026-10-21"
   },
   {
-    "dataset_id": "DS-2026-MAPPING-ASSAYED-SAMPLES",
-    "dataset_name": "Expression samples with mutation calls",
-    "version": "v1 / 2026-09-15;",
+    "dataset_id": "DS-2026-MAPPING-TCGA-CLINICAL",
+    "dataset_name": "TCGA OncoTree and MSI annotations",
+    "version": "v2 / 2026-10-21",
     "dataset_type": "benchmark",
     "related_ai_system": [
       "Ingenix Research Cluster"
     ],
-    "intended_purpose": "Distinguish samples with mutation data but no qualifying mutation from samples without mutation calls.",
-    "source": "TCGA Pan-Cancer Atlas 2018 mutation files distributed through cBioPortal Datahub (https://github.com/cBioPortal/datahub/tree/master/public); Gustave Roussy META-PRISM publication supplementary Table S6 (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/#sec63); Broad Institute DepMap Public 25Q3, OmicsSomaticMutations.csv (https://depmap.org/portal/download/all/).",
-    "source_and_acquisition": "TCGA: The preparation pipeline downloaded cBioPortal Datahub study folders with Git LFS and read public/*tcga_pan_can_atlas_2018*/data_mutations.txt. Cohort mutation tables were concatenated and their variant annotations were standardized. META-PRISM: Mutation calls were read from Table S6 of the publication supplement. DepMap: Raw somatic mutation calls were read from OmicsSomaticMutations.csv for release 25Q3; download URLs were resolved through the provider endpoint https://depmap.org/portal/api/download/files. For the benchmark, variants with vep_impact HIGH or MODERATE were collapsed to gene-level mutation presence. TCGA patient-level calls and META-PRISM subject-level calls were linked to the corresponding expression samples, while DepMap models were matched by Consensus_model_id. The derived exports were aligned to the expression sample identifiers.",
-    "license": "TCGAl: GDC open-access. META-PRISM: (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/). DepMap: CC by 4.0",
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "TCGA Pan-Cancer Atlas 2018, cBioPortal datahub; https://github.com/cBioPortal/datahub/tree/master/public",
+    "source_and_acquisition": "Clinical annotations and MSI scores originate from TCGA clinical tables. OncoTree codes were harmonized upstream.",
+    "license": "ODbL 1.0 with cBioPortal attribution. The datahub also includes a TCGA-specific GDAC notice; applicable study terms require release approval. https://github.com/cBioPortal/datahub#license",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "One row per dataset and expression sample represented in the mutation-call matrices, including all-zero rows. The supplied reference export contains 10696 rows: tcga 8877, depmap 1465, metaprism 354.",
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
     "legal_basis_and_rights": "n/a",
-    "quality_criteria": "Verified column names and string types; no nulls, empty key values or duplicate keys. All assayed samples exist in the corresponding expression H5ADs.",
+    "quality_criteria": "10,953 rows and exactly 4 columns: sample_id, cancer_type_oncotree, msi_score_mantis, msi_score_sensor. Row count matches the specification. Parquet equality was checked against the selected source values.",
     "processing_steps": [
-      "Sample selection: All sample rows were taken from the documented mutation matrices. Samples whose variants did not pass the non-silent filter were included.",
-      "Sample matching: TCGA patients and META-PRISM subjects were associated with their expression samples. DepMap models were matched to expression sample identifiers.",
-      "Sample export: Unique dataset and sample_id pairs were exported for samples present in the expression datasets."
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only sample_id, cancer_type_oncotree, msi_score_mantis, msi_score_sensor were retained.",
+      "Row and value selection: All source rows were retained.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
     ],
-    "bias_assessment": "TCGA, META-PRISM and DepMap differ in biological models and mutation ascertainment.",
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
     "data_owner": "Alicja Augustyniak",
-    "last_reviewed": "2026-09-15"
+    "last_reviewed": "2026-10-21"
   },
   {
-    "dataset_id": "DS-2026-MAPPING-ASSAYED-GENES",
-    "dataset_name": "Gene sets represented in the mutation matrices",
-    "version": "v1 / 2026-09-15;",
+    "dataset_id": "DS-2026-MAPPING-DEPMAP-MUTATIONS",
+    "dataset_name": "DepMap 25Q3 somatic variant calls",
+    "version": "v2 / 2026-10-21",
     "dataset_type": "benchmark",
     "related_ai_system": [
       "Ingenix Research Cluster"
     ],
-    "intended_purpose": "Define the per-dataset gene sets and their 124-gene intersection for mutation-based evaluation.",
-    "source": "TCGA Pan-Cancer Atlas 2018 mutation files distributed through cBioPortal Datahub (https://github.com/cBioPortal/datahub/tree/master/public); Gustave Roussy META-PRISM publication supplementary Table S6 (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/#sec63); Broad Institute DepMap Public 25Q3, OmicsSomaticMutations.csv (https://depmap.org/portal/download/all/).",
-    "source_and_acquisition": "TCGA: The preparation pipeline downloaded cBioPortal Datahub study folders with Git LFS and read public/*tcga_pan_can_atlas_2018*/data_mutations.txt. Cohort mutation tables were concatenated and their variant annotations were standardized. META-PRISM: Mutation calls were read from Table S6 of the publication supplement. DepMap: Raw somatic mutation calls were read from OmicsSomaticMutations.csv for release 25Q3; download URLs were resolved through the provider endpoint https://depmap.org/portal/api/download/files. For the benchmark, variants with vep_impact HIGH or MODERATE were collapsed to gene-level mutation presence. TCGA patient-level calls and META-PRISM subject-level calls were linked to the corresponding expression samples, while DepMap models were matched by Consensus_model_id. The derived exports were aligned to the expression sample identifiers.",
-    "license": "TCGA: GDC open-access. META-PRISM: (https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/). DepMap: CC by 4.0",
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "Broad Institute, DepMap Public 25Q3; https://depmap.org/portal/download/all/",
+    "source_and_acquisition": "Calls originate from OmicsSomaticMutations.csv. Model and gene identifiers were harmonized upstream.",
+    "license": "CC BY 4.0. https://creativecommons.org/licenses/by/4.0/",
     "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "One row per dataset and Ensembl gene represented as a column in the mutation matrix. The supplied reference export contains 37298 rows: tcga 18566, depmap 18608, metaprism 124.",
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
     "legal_basis_and_rights": "n/a",
-    "quality_criteria": "Verified column names and string types; no nulls, empty key values or duplicate keys. All assayed samples exist in the corresponding expression H5ADs.",
+    "quality_criteria": "750,004 rows and exactly 9 columns: Consensus_model_id, is_default_entry_for_model, gene_id_ensembl, protein_change, variant_info, ref, alt, gt, ps. Row count matches the specification. Parquet equality was checked against the selected source values.",
     "processing_steps": [
-      "Gene extraction: Gene columns were extracted from the documented TCGA, DepMap and META-PRISM mutation matrices.",
-      "Gene export: Unique dataset and gene_id_ensembl pairs were exported. The gene sets were not restricted to the expression gene panel.",
-      "Comparison panel: The intersection of the three gene sets was used as the comparison panel. The supplied intersection contained 124 genes."
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only Consensus_model_id, is_default_entry_for_model, gene_id_ensembl, protein_change, variant_info, ref, alt, gt, ps were retained.",
+      "Row and value selection: All rows and genes were retained. The is_default_entry_for_model flag was preserved; every source row is a default entry.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
     ],
-    "bias_assessment": "TCGA, META-PRISM and DepMap differ in biological models and mutation ascertainment.",
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
     "data_owner": "Alicja Augustyniak",
-    "last_reviewed": "2026-09-15"
+    "last_reviewed": "2026-10-21"
   },
   {
-    "dataset_id": "DS-2026-MAPPING-MONDO",
-    "dataset_name": "MONDO disease ontology",
-    "version": "v1 / 2026-09-14",
+    "dataset_id": "DS-2026-MAPPING-DEPMAP-SUBTYPES",
+    "dataset_name": "DepMap 25Q3 OncoTree and MSI subtype tags",
+    "version": "v2 / 2026-10-21",
     "dataset_type": "benchmark",
     "related_ai_system": [
       "Ingenix Research Cluster"
     ],
-    "intended_purpose": "Provide disease identifiers, names and ancestor relationships for ontology-based comparison of disease labels in the mapping benchmark.",
-    "source": "Mondo Disease Ontology (MONDO), Monarch Initiative. Official ontology reference: https://purl.obolibrary.org/obo/mondo.obo.",
-    "source_and_acquisition": "Only the MONDO component was retained in the benchmark export. The official MONDO ontology is available in OWL, OBO and JSON formats.",
-    "license": "Creative Commons Attribution 4.0 International (CC BY 4.0). MONDO licence: https://github.com/monarch-initiative/mondo/blob/master/LICENSE.",
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "Broad Institute, DepMap Public 25Q3; https://depmap.org/portal/download/all/",
+    "source_and_acquisition": "Subtype annotations were represented upstream as model-to-subtype records.",
+    "license": "CC BY 4.0. https://creativecommons.org/licenses/by/4.0/",
+    "is_synthetic": false,
     "ai_augmentation": null,
-    "provenance": "The export contains 15,481 MONDO nodes, 25,555 parent relations and ten hierarchy columns. The non-human animal disease branch was excluded.",
-    "legal_basis_and_rights": "The export contains ontology identifiers, disease names and graph relationships, without patient or sample records. Reuse of MONDO ontology content is subject to CC BY 4.0",
-    "quality_criteria": "The recorded export audit confirms an acyclic graph with one root, MONDO:0000001, and preserved reachability among retained MONDO nodes. No dangling source parent references or conflicting nonempty source MONDO mappings were recorded.",
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "5,828 rows and exactly 2 columns: Consensus_model_id, Subtype. Row count matches the specification. Parquet equality was checked against the selected source values.",
     "processing_steps": [
-      "Source preparation: MONDO terms and relationships were prepared for the benchmark.",
-      "Source validation: The source graph was checked for cycles and dangling parent references. Nonempty MONDO identifier mappings were checked for conflicts.",
-      "Hierarchy processing: Direct parent relations between retained MONDO nodes were preserved. Paths through non-MONDO nodes were projected onto the nearest MONDO ancestors. Nine derived relations and their source paths were recorded.",
-      "Branch filtering: MONDO:0005583 and its 62 exclusive descendants were excluded. No benchmark sample carried an excluded label.",
-      "Hierarchy annotation: Parent relationships, ancestor closures, names, counts and levels were recomputed. The mondo_level field was defined as the minimum distance from a root. The original scalar parent was retained when it remained a direct parent. Otherwise, the first sorted parent was selected. The direct_ancestors_ids field was retained as the authoritative multi-parent representation.",
-      "Export schema: Ten columns were retained, covering MONDO identifiers, names, levels, parents and ancestors."
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only Consensus_model_id, Subtype were retained.",
+      "Row and value selection: Rows were retained only when Subtype matched an exported OncoTree code or the MSI tag. Other molecular tags were excluded.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
     ],
-    "bias_assessment": "Coverage and hierarchy reflect the source ontology content and the selected MONDO subgraph.",
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
     "data_owner": "Alicja Augustyniak",
-    "last_reviewed": "2026-09-16"
+    "last_reviewed": "2026-10-21"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-DEPMAP-COPY-NUMBER",
+    "dataset_name": "DepMap 25Q3 copy-number vectors",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "Broad Institute, DepMap Public 25Q3; https://depmap.org/portal/download/all/",
+    "source_and_acquisition": "Copy-number measurements were represented upstream as ordered per-model vectors.",
+    "license": "CC BY 4.0. https://creativecommons.org/licenses/by/4.0/",
+    "is_synthetic": false,
+    "ai_augmentation": null,
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "1,095 rows and exactly 2 columns: Consensus_model_id, omics_features. Row count matches the specification. Parquet equality was checked against the selected source values.",
+    "processing_steps": [
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only Consensus_model_id, omics_features were retained.",
+      "Row and value selection: All model rows and feature positions were retained. Float64 precision and feature order were preserved.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
+    ],
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-10-21"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-DEPMAP-FEATURES",
+    "dataset_name": "DepMap 25Q3 copy-number feature identifiers",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "Broad Institute, DepMap Public 25Q3; https://depmap.org/portal/download/all/",
+    "source_and_acquisition": "Feature identifiers accompany the copy-number matrix.",
+    "license": "CC BY 4.0. https://creativecommons.org/licenses/by/4.0/",
+    "is_synthetic": false,
+    "ai_augmentation": null,
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "1 rows and exactly 1 columns: copy_number_features. Row count matches the specification. Parquet equality was checked against the selected source values.",
+    "processing_steps": [
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only copy_number_features were retained.",
+      "Row and value selection: The single row and complete ordered copy_number_features vector were retained.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
+    ],
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-10-21"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-ONCOTREE",
+    "dataset_name": "OncoTree cancer codes and levels",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "OncoTree; https://oncotree.mskcc.org/; https://github.com/cBioPortal/oncotree",
+    "source_and_acquisition": "All 897 exported OncoTree codes and levels match the official release oncotree_2025_10_03 (3 October 2025).",
+    "license": "CC BY 4.0. https://github.com/cBioPortal/oncotree/blob/master/LICENSE.md",
+    "is_synthetic": false,
+    "ai_augmentation": null,
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "897 rows and exactly 2 columns: oncotree_code, oncotree_level. Row count matches the specification. Parquet equality was checked against the selected source values.",
+    "processing_steps": [
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only oncotree_code, oncotree_level were retained.",
+      "Row and value selection: All codes and levels were retained.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
+    ],
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-10-21"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-CELL-LINES",
+    "dataset_name": "Cellosaurus MSI and mutation annotations for DepMap lines",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "SIB Swiss Institute of Bioinformatics, Cellosaurus; https://www.cellosaurus.org/; Bairoch (2018), doi:10.7171/jbt.18-2902-002",
+    "source_and_acquisition": "MSI and mutation annotations originate from Cellosaurus.",
+    "license": "CC BY 4.0. https://www.cellosaurus.org/description.html",
+    "is_synthetic": false,
+    "ai_augmentation": null,
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "1,507 rows and exactly 3 columns: cvcl_id, msi_status, mutations. Row count matches the specification. Parquet round-trip equality was checked against the selected source values.",
+    "processing_steps": [
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only cvcl_id, msi_status, mutations were retained.",
+      "Row and value selection: Rows were retained when cvcl_id occurred in DepMap expression file.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
+    ],
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-10-21"
+  },
+  {
+    "dataset_id": "DS-2026-MAPPING-METAPRISM-CLINICAL",
+    "dataset_name": "META-PRISM patient, sample and OncoTree annotations",
+    "version": "v2 / 2026-10-21",
+    "dataset_type": "benchmark",
+    "related_ai_system": [
+      "Ingenix Research Cluster"
+    ],
+    "intended_purpose": "Research benchmarking of mutation-axis matching between patient tumours and biological models. This table supplies evidence for benchmark labels; it is not a clinical decision system.",
+    "source": "Gustave Roussy, META-PRISM; Pradat et al. (2023), doi:10.1158/2159-8290.CD-22-0966; https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/",
+    "source_and_acquisition": "Clinical records originate from supplementary Table S1. OncoTree codes were harmonized upstream.",
+    "license": "The publication states CC BY-NC-ND 4.0. https://pmc.ncbi.nlm.nih.gov/articles/PMC10157368/",
+    "is_synthetic": false,
+    "ai_augmentation": null,
+    "provenance": "The existing processed source table was reduced to the columns and rows required by the mutation-axis benchmark. Original provider files were not reprocessed during this migration.",
+    "legal_basis_and_rights": "n/a",
+    "quality_criteria": "1,031 rows and exactly 3 columns: patient_id, sample_id, cancer_type_oncotree. Row count matches the specification. Parquet round-trip equality was checked against the selected source values.",
+    "processing_steps": [
+      "Input preparation: Original annotations were preserved.",
+      "Column selection: Only patient_id, sample_id, cancer_type_oncotree were retained.",
+      "Row and value selection: All patient rows were retained. The full mutation MAF remains external.",
+      "Export: The index was reset. The table was written as Parquet with Zstandard compression, without a stored index. No gene-panel filter was applied."
+    ],
+    "bias_assessment": "Coverage and missing annotations reflect the source study or resource. Absence of a mutation record alone does not establish a negative biomarker label.",
+    "data_owner": "Alicja Augustyniak",
+    "last_reviewed": "2026-10-21"
   }
 ]

mutations/{assayed_genes.parquet → cell_lines.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d33fa36a3d14dad9e6e752e29dfe1a7d3e2765014b14a406410ff60b675e0b6
-size 178238
+oid sha256:61c91ad791d6e20078c0320a380c93b69dc8b500d5f3931b57f2d756787d05da
+size 47569

mutations/depmap_copy_number.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13a96cc6816f82e562fb74368c072868670193b147608c7fc38cc160d1464af0
+size 156251505

mutations/{assayed_samples.parquet → depmap_features.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ffe9d8a4a1d7b01e0778261d6d4958ef702c7d90d06344af01effa8871076da
-size 100485
+oid sha256:927187c3c1e69130cde9bb1ca4af260f6b11d0eabfb4c791fc29a3c2db283d9a
+size 106102

mutations/{mutated_genes.parquet → depmap_mutations.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:582224d0f3702a6e20edbb9a035690450f9ddd11e3fbf43e8fc4655f616f8f42
-size 1666686
+oid sha256:9c63e32372134899180a52a1d91e998bdba15e2683fbe9a18cc5fb5e532cea5f
+size 6205882

mutations/depmap_subtypes.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b5e67d76477ec93b799bd19b07876649bf4687119a605db3fb40aff887f9fe8
+size 16415

mutations/metaprism_clinical.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af4bb17a9fdaf1ead96ef470bebc0b8a96de94afc1d3c17c1dfff9cffe330da7
+size 7983

mutations/oncotree.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99fecfd1db1a43f66b9f7d441814ad6e86167089d17d8c4954376dfbdfdbd116
+size 6190

mutations/tcga_clinical.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:672f2e541dfa29c7f6eb2ebea4c0902cc9e83a04d50735ba5176de1eaa5fea77
+size 80064

mutations/tcga_mutations.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f18e212edaea927b93b1a2bc2333bd27a1c14f7257babbc20f555e715264d713
+size 16919524