34:["$","title",null,{"children":"Diseasome: PGS/23andme SNP Overlap Analysis (v2–v5) | lorenasandoval88 | Observable"}] 3c:T283b,

1,759 PGS models ≤1,000 variants/model 4 chips · GRCh37

Average SNP overlap

Mean percentage of each model’s SNPs on the chip

28.21%

41.29%

32.13%

31.48%

v3 · Highest average coverage

Models with high overlap

■ ≥80% overlap ■ ≥95% overlap

31 10

71 27

39 15

121 32

v5 · Most high-overlap models

The ≥80% count includes all models with ≥95% overlap. SNP coverage ≠ predictive accuracy.

3d:T498,Inputs.table(filteredOverlap, { columns: [ "pgsId", "name", "trait", "variantsInFile", "v2_overlapPercent", "v3_overlapPercent", "v4_overlapPercent", "v5_overlapPercent", "v5_matched", "v5_missing" ], header: { pgsId: "PGS ID", name: "Model", trait: "Trait", variantsInFile: "SNPs", v2_overlapPercent: "v2 %", v3_overlapPercent: "v3 %", v4_overlapPercent: "v4 %", v5_overlapPercent: "v5 %", v5_matched: "Matched", v5_missing: "Missing" }, width: { pgsId: 95, name: 115, trait: 110, variantsInFile: 60, v2_overlapPercent: 65, v3_overlapPercent: 65, v4_overlapPercent: 65, v5_overlapPercent: 65, v5_matched: 75, v5_missing: 75 }, format: { trait: d => d?.length > 20 ? d.slice(0, 20) + "…" : d, v2_overlapPercent: d => d == null ? "" : `${Number(d).toFixed(1)}%`, v3_overlapPercent: d => d == null ? "" : `${Number(d).toFixed(1)}%`, v4_overlapPercent: d => d == null ? "" : `${Number(d).toFixed(1)}%`, v5_overlapPercent: d => d == null ? "" : `${Number(d).toFixed(1)}%` }, rows: 10, maxWidth: 950, select: false })3e:T1e7b,## Inputs | Source | Location | |---|---| | PGS scoring files | `gs://allpgs/scores/GRCh37/` | | v2 manifest | `gs://all23_0710/chip_manifests/v2_build37_95pct/` | | v3 manifest | `gs://all23_0710/chip_manifests/v3_build37_95pct/` | | v4 manifest | `gs://all23_0710/chip_manifests/v4_build37_95pct/` | | v5 manifest | `gs://all23_0710/chip_manifests/v5_build37_95pct/` | Each chip directory contains `manifest.json` and `variant_manifest_core.json`. **Core definition:** SNP coordinates present in ≥95% of the sampled files for each chip. ## Processing 1. Load v2–v5 GRCh37 core manifests. 2. Identify PGS models with ≤1,000 scoring-file rows. 3. Parse harmonized PGS scoring files. 4. Match SNPs by `chromosome:position`. 5. Deduplicate coordinates and validate variant records. 6. Calculate direct overlap for each chip (0–100%). 7. Generate model-level reports, summary statistics, and QC results. **Matching:** GRCh37 coordinates are primary; rsIDs are retained for diagnostics. ## Results ### 1. 23andMe Chip Manifest Coverage GRCh37 chip manifests were generated using SNP positions present in at least 95% of the sampled files for each chip version. | Chip | 23andMe files | Core SNP positions | |---|---:|---:| | v2 | 3 | 544,839 | | **v3** | **17** | **959,702** | | v4 | 74 | 596,247 | | v5 | 64 | 622,753 | **Finding:** v3 contains the largest number of SNP positions, followed by v5, v4, and v2. However, a larger chip does not necessarily provide better coverage for every PGS model. ### 2. PGS Model Overlap A total of **1,759 PGS Catalog models containing ≤1,000 variants** were compared against the four chip manifests. Each model's overlap was calculated as the percentage of its SNPs directly represented on each chip. | Metric | v2 | v3 | v4 | v5 | |---|---:|---:|---:|---:| | Mean model overlap | 28.21% | **41.29%** | 32.13% | 31.48% | | Models with ≥80% overlap | 31 | 71 | 39 | **121** | | Models with ≥95% overlap | 10 | 27 | 15 | **32** | | Models with 100% overlap | 10 | **24** | 13 | 22 | **Finding:** v3 has the highest average overlap across the 1,759 models, while v5 has the largest number of individual models exceeding the 80% and 95% overlap thresholds. These results measure different aspects of coverage and should not be interpreted as overall superiority of one chip. ### 3. Coverage of All Unique PGS SNPs To evaluate broader genomic coverage, SNPs from the 1,759 models were combined and deduplicated. - Total scoring-file variant entries: 379,350 - Valid chromosome-position entries: 378,648 - **Unique genomic target positions: 178,278** Each chip was compared against these 178,278 unique PGS target positions. | Chip | Observed PGS SNPs | Missing PGS SNPs | Coverage | |---|---:|---:|---:| | v2 | 43,042 | 135,236 | 24.14% | | **v3** | **64,196** | **114,082** | **36.01%** | | v4 | 44,202 | 134,076 | 24.79% | | v5 | 26,683 | 151,595 | 14.97% | **Finding:** v3 provides the broadest coverage of unique PGS target positions, whereas v5 contains the fewest of these positions. This supports using imputation to recover missing PGS variants, particularly for models with limited direct coverage. ### 4. Why Does v5 Have More Models With ≥80% Overlap? Although v5 covers fewer unique PGS SNPs overall, it has more models passing the ≥80% threshold. This difference is largely associated with the size and SNP composition of individual PGS models. Among the 121 v5 models with ≥80% overlap: | Variants per model | Number of models | |---|---:| | 1–20 | 70 | | 21–50 | 19 | | 51–100 | 24 | | 101–200 | 6 | | 201–500 | 2 | | 501–1,000 | 0 | | **Total** | **121** | - **70 of the 121 models (57.9%) contain only 1–20 variants.** - Only 2 models contain more than 200 variants. - Some PGS models share identical or highly overlapping SNP sets, so the 121 models do not necessarily represent 121 independent SNP panels. - v5 appears particularly well matched to certain small PGS models rather than providing consistently better coverage across all models. For example, breast cancer model PGS000001 (PRS77) has: | Chip | Matched SNPs | Overlap | |---|---:|---:| | v3 | 54 of 77 | 70.1% | | v5 | 74 of 77 | 96.1% | However, breast cancer PRS313 (PGS000004) has only approximately 22.4% direct overlap on v5. This illustrates why chip-specific overlap varies substantially between PGS models. ### 5. Interpretation and Key Findings 1. **v3 provides the broadest genomic coverage.** It contains the most core SNP positions, has the highest mean model overlap (41.29%), and covers the greatest proportion of unique PGS targets (36.01%). 2. **v5 performs particularly well for a subset of small PGS models.** It has the most models reaching ≥80% overlap (121), but most of these models contain relatively few variants. 3. **PGS model size and SNP selection matter.** A chip with fewer overall SNPs can achieve high coverage for a particular PGS if it includes most of that model's required variants. 4. **The relationship between chip version and PGS publication year has not been established.** Additional analysis of PGS release dates would be needed to determine whether v5 aligns better with newer models. 5. **Most unique PGS target SNPs are absent from every chip.** This motivates extending FastImpute to predict missing variants rather than relying exclusively on directly genotyped SNPs. 6. **Direct SNP overlap is not equivalent to PGS accuracy.** The current comparisons use GRCh37 chromosome-position matching. Allele compatibility, genotype availability, imputation accuracy, and downstream PGS agreement require separate validation. **Overall conclusion:** The v3 chip offers broader direct coverage of PGS-associated genomic positions, whereas v5 provides particularly high coverage for a subset of small PGS models. These findings demonstrate that direct PGS applicability depends on both chip design and model composition, supporting chip-specific imputation and validation in Diseasome. ## Outputs ### PGS Model Overlap Analysis ```text gs://all23_0710/pgs_overlap_v2_v5_95pct/ ├── README.md ├── summary.json ├── results/ │ ├── overlap_v2_0_100.json │ ├── overlap_v3_0_100.json │ ├── overlap_v4_0_100.json │ ├── overlap_v5_0_100.json │ └── model_chip_overlap.csv └── qc/ ├── validation_summary.json └── failed_pgs.json ``` ### PGS Target Union for FastImpute ```text gs://allpgs/fastImp/ ├── summary.json ├── model_chip_overlap.csv ├── unique_pgs_positions.json ├── unique_pgs_variants_allele_aware.json ├── v2_missing_targets.json ├── v3_missing_targets.json ├── v4_missing_targets.json └── v5_missing_targets.json ``` **Next step:** Use the missing-target lists to extend FastImpute beyond PRS313, beginning with v5 and validating predicted genotypes and downstream PGS estimates against reference genotypes and available imputation results. ## Diseasome Tab 2 Integration Use precomputed overlap results to support: - Filter by chip version (v2, v3, v4, v5). - Filter by PGS model size (≤1,000 variants). - Filter by SNP overlap threshold (≥80%, ≥90%, ≥95%, 100%). - Display matched/missing variant counts and QC information. - Join PGS Catalog metadata for model names, traits, publications, and licenses. - Identify models requiring FastImpute or other imputation methods. **Next:** Combine the overlap reports with PGS Catalog metadata into a browser-ready model-selection dataset. **Note:** The 95% chip-core presence threshold is separate from the 0–100% PGS model-overlap percentage. These results replace the earlier 20-file v4/v5 analysis and should not be interpreted as imputation or PRS validation.3f:[{"id":"8ed75133c1f546a234c71e33601c9feb4f6e0c5e2f22e432a210dcaecc5079124a7300e0d39fcfd9d6c0292bb687d357cd65df5b3657f376f6f6983bf3c5cd9b","href":"https://static.observableusercontent.com/files/8ed75133c1f546a234c71e33601c9feb4f6e0c5e2f22e432a210dcaecc5079124a7300e0d39fcfd9d6c0292bb687d357cd65df5b3657f376f6f6983bf3c5cd9b","name":"pgs_overlap_v2_v5_95pct_results_model_chip_overlap.csv","type":"text/csv","lastModified":1791488815582,"size":271728},{"id":"6be4b3af1c1b0d6e2182f3c4d35ae625afe8648418e6102eeb302ebcb65957ff5f805044c1f5e8525b0301c35662da43f6eafb4d017f7ab0605b977c7d0014e7","href":"https://static.observableusercontent.com/files/6be4b3af1c1b0d6e2182f3c4d35ae625afe8648418e6102eeb302ebcb65957ff5f805044c1f5e8525b0301c35662da43f6eafb4d017f7ab0605b977c7d0014e7","name":"overlap_v2_0_100.json","type":"application/json","lastModified":1791489488943,"size":1688091},{"id":"aac7d9f8f5f837dfda10c16811e1a5c04a9290aca415df0208b7535992b71f8287072f1e032dc86f1ad2c978f0d2fa550e09e78c41b981331feda5f07eaf5a3a","href":"https://static.observableusercontent.com/files/aac7d9f8f5f837dfda10c16811e1a5c04a9290aca415df0208b7535992b71f8287072f1e032dc86f1ad2c978f0d2fa550e09e78c41b981331feda5f07eaf5a3a","name":"overlap_v3_0_100.json","type":"application/json","lastModified":1791489488961,"size":1688668},{"id":"9c9230c29781620d06a7533eeebd22b6d769cde485b5739d22fe133cfa6186155a18730a5fb965f83d1941e764c9419fd81bc7aa2e76bac67fd333e9a897c73a","href":"https://static.observableusercontent.com/files/9c9230c29781620d06a7533eeebd22b6d769cde485b5739d22fe133cfa6186155a18730a5fb965f83d1941e764c9419fd81bc7aa2e76bac67fd333e9a897c73a","name":"overlap_v4_0_100.json","type":"application/json","lastModified":1791489488977,"size":1688444},{"id":"524adc97325745c27a261ede99351bf0435226a15e4910db4c1a4ba92e06a7c0dfa7b2c1033424a8a0940992e1e8d428633825213b08ea5486ff6096b00df147","href":"https://static.observableusercontent.com/files/524adc97325745c27a261ede99351bf0435226a15e4910db4c1a4ba92e06a7c0dfa7b2c1033424a8a0940992e1e8d428633825213b08ea5486ff6096b00df147","name":"overlap_v5_0_100.json","type":"application/json","lastModified":1791489488994,"size":1688201},{"id":"8ed75133c1f546a234c71e33601c9feb4f6e0c5e2f22e432a210dcaecc5079124a7300e0d39fcfd9d6c0292bb687d357cd65df5b3657f376f6f6983bf3c5cd9b","href":"https://static.observableusercontent.com/files/8ed75133c1f546a234c71e33601c9feb4f6e0c5e2f22e432a210dcaecc5079124a7300e0d39fcfd9d6c0292bb687d357cd65df5b3657f376f6f6983bf3c5cd9b","name":"model_chip_overlap.csv","type":"text/csv","lastModified":1791489595952,"size":271728},{"id":"83cfea8a7cb10ba8f7a410973fb3c3300de0cbef67a0569a71411b6c01425a20bdf699628f889fb15b3797933093413dc3934a4efd3f6ab146e249f04ac539be","href":"https://static.observableusercontent.com/files/83cfea8a7cb10ba8f7a410973fb3c3300de0cbef67a0569a71411b6c01425a20bdf699628f889fb15b3797933093413dc3934a4efd3f6ab146e249f04ac539be","name":"Dockerfile.txt","type":"text/plain","lastModified":1791490644977,"size":126},{"id":"94eaa0705018ce05e39ba2070962b31f72fab03daec98bb1780a643a26446404854cfdaeee1d2fed0cbb9e5dfc9867b21a6c0e2c566e370e025a96842a55ae1e","href":"https://static.observableusercontent.com/files/94eaa0705018ce05e39ba2070962b31f72fab03daec98bb1780a643a26446404854cfdaeee1d2fed0cbb9e5dfc9867b21a6c0e2c566e370e025a96842a55ae1e","name":"index.js","type":"text/javascript","lastModified":1791490532432,"size":26425},{"id":"9b68e25c88c2c1cbe1f32f14b7fc4edaf2fbf906c437e9c195b29f5ce3dd05595edaf1e0f3a9f2b921253c2d50143abd1762b25c323fc9824087a8948cf5f46e","href":"https://static.observableusercontent.com/files/9b68e25c88c2c1cbe1f32f14b7fc4edaf2fbf906c437e9c195b29f5ce3dd05595edaf1e0f3a9f2b921253c2d50143abd1762b25c323fc9824087a8948cf5f46e","name":"package.json","type":"application/json","lastModified":1791490584490,"size":382},{"id":"2f672ba567c8be1171a17a41ea6c2ea90b3bdd9438a6d629425b8b632165fcaeb5f0aa40cb817fe0d8059cf0777227b94e3e2a389ab77016f72ded0b29e72bf1","href":"https://static.observableusercontent.com/files/2f672ba567c8be1171a17a41ea6c2ea90b3bdd9438a6d629425b8b632165fcaeb5f0aa40cb817fe0d8059cf0777227b94e3e2a389ab77016f72ded0b29e72bf1","name":"README.md","type":"text/markdown","lastModified":1791490571787,"size":9101},{"id":"aff50890d8d23ba753925346ab6415e27495265b709f2c6181f5313046c67f60cf412ff6c610b0b65cda968dedc55bb72b7ee30d44ec7a7a2b393b75b2eb6976","href":"https://static.observableusercontent.com/files/aff50890d8d23ba753925346ab6415e27495265b709f2c6181f5313046c67f60cf412ff6c610b0b65cda968dedc55bb72b7ee30d44ec7a7a2b393b75b2eb6976","name":"pgs_overlap_flow.png","type":"image/png","lastModified":1791492497782,"size":1533099}] 38:{"id":"f6f7c0aa9945f2ff","slug":"pgs-models-overlap-with-23andme-chips","title":"Diseasome: PGS/23andme SNP Overlap Analysis (v2–v5)","thumbnail":"c388b7d9fdfd41bd27c5773c0854cb09e9646a28ba3ac2001c0f47c156d30c2a","access_level":"public","create_time":"$D2026-08-03T15:26:25.064Z","update_time":"$D2026-10-16T16:17:24.771Z","view_time":null,"first_public_time":"$D2026-08-03T15:26:25.064Z","has_unpublished_edits":null,"fork":false,"stars":0,"starred":false,"owner":{"type":"team","id":"5a5147324ed15c62","login":"lorenasandoval88","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":"starter_2024","bio":"","followed_by_you":false,"follows_you":false},"version":239,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"5c479a2c5ae5f52f","login":"lorenasandoval","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":null,"bio":"","followed_by_you":false,"follows_you":false}],"members":[{"type":"individual","id":"5c479a2c5ae5f52f","login":"lorenasandoval","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":null,"role":"editor"},{"type":"team","id":"5a5147324ed15c62","login":"lorenasandoval88","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":"starter_2024","role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":0,"collection":{"id":"8bdfe82d7003e3b0","slug":"imputation","title":"imputation","owner":{"id":"5a5147324ed15c62","login":"lorenasandoval88","name":""}},"type":"notebook","body":{"titling":"auto","title":"Diseasome: PGS/23andme SNP Overlap Analysis (v2–v5)","cells":[{"id":0,"value":"# Diseasome: PGS/23andme SNP Overlap Analysis (v2–v5)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"},{"id":65,"value":"$3c","pinned":false,"hidden":false,"mode":"html","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":71,"value":"","pinned":true,"hidden":false,"mode":"ts","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":33,"value":"model_23chip_overlap = await FileAttachment(\"model_chip_overlap.csv\").csv({typed: true})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":43,"value":"viewof selectedTrait = Inputs.select(\n [\n \"All traits\",\n ...new Set(model_23chip_overlap.map(d => d.trait).filter(Boolean))\n ].map(String).sort((a, b) =>\n a === \"All traits\" ? -1 :\n b === \"All traits\" ? 1 :\n a.localeCompare(b)\n ),\n {\n label: \"Filter by trait\",\n value: \"All traits\"\n }\n)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":45,"value":"viewof modelSearch = Inputs.text({\n label: \"Search PGS models\",\n placeholder: \"Type PGS ID, model name, or trait...\",\n submit: false\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":46,"value":"filteredOverlap = model_23chip_overlap.filter(d => {\n const traitMatch =\n selectedTrait === \"All traits\" ||\n d.trait === selectedTrait;\n\n const q = modelSearch.trim().toLowerCase();\n\n const searchMatch = [\n d.pgsId,\n d.name,\n d.trait\n ].some(v => String(v ?? \"\").toLowerCase().includes(q));\n\n return traitMatch && searchMatch;\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":34,"value":"$3d","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":26,"value":"v5 = FileAttachment(\"overlap_v5_0_100.json\").json()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":31,"value":"v4 = FileAttachment(\"overlap_v4_0_100.json\").json()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":57,"value":"v3 = FileAttachment(\"overlap_v3_0_100.json\").json()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":59,"value":"v2 = FileAttachment(\"overlap_v2_0_100.json\").json()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":16,"value":"$3e","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"","database":"$undefined"}],"files":"$W3f","stdlib":"1"}} 33:null