New variant interpretation guidelines from ACMG-AMP for SNVs, SVC v4.0, is expected in 2024. This version introduces several significant changes compared to the existing 2015 version and requires adaptation both in the ELLA UI and ACMG rules engine. ## Background Compared to the existing 2015 guidelines (v3.0), v4.0 introduces several significant changes, most importantly: - Existing evidence codes (PVS1, BS1, BS2, etc.) are replaced with descriptive codes with separate weights - Weights are given as points based on a Bayesian framework ([Tavtigian et al. 2020](https://pubmed.ncbi.nlm.nih.gov/32720330/)). - Detailed decision trees are introduced to guide point selection on individual criteria - Class 3-variants are split into three sub-categories (high-mid-low) - Incorporates gene-disease validity The framework aligns more closely with the 2020 CNV guidelines (see related issues alleles/ella#263 and alleles/ella#264), which will also get an update to further unify the guidelines for SNVs and CNVs. ### Resources - Presentation (Zoom/YouTube): [Overview of ACMG/AMP v4 Sequence Variant Guidelines](https://www.youtube.com/watch?v=4NqVGcPy050) - Various documents (Google Drive): [ACMG_SVC_v4](https://drive.google.com/drive/folders/1r0DribTDCmZf-zLLDsYdWGPwuawgrLyL) - Reference: Fitting a naturally scaled point system to the ACMG/AMP variant classification guidelines ([Tavtigian et al. 2020](https://pubmed.ncbi.nlm.nih.gov/32720330/)) - SVI General Recommendations for Using ACMG/AMP Criteria (addendums to SVC 3.0 that will be incorporated in SVC 4.0): [ClinGen web page](https://www.clinicalgenome.org/working-groups/sequence-variant-interpretation) ## Detailed background <details><summary>Click to expand</summary> ### New evidence codes New codes are descriptive of type of evidence but not of weight/strength. The latter will instead be signified by point value combined with the code, e.g. CLN_DNV_+4 or POP_FRQ_-3 Each code will have decision trees for deciding point values within a defined range that is unique to each code. **NOTE**: Evidence codes given below are draft codes and are subject to change. #### Population (POP) Code | Description ------ | ------ POP_FRQ | Population Freq. #### Clinical evidence (CLN) Code | Description ------ | ------ CLN_CCR | Case:control ratio CLN_COB | Case observation counts CLN_SEG | Segregation data CLN_PHE | Specific Phenotype CLN_DNV | De novo #### Molecular impact (IMP) Code | Description ------ | ------ ? IMP_NMD | Predicted to undergo NMD ? IMP_NUL | Absent protein. Referred to in some slides, alternative to IMP_NMD? IMP_CDS | Impact to coding sequence IMP_MSS | Missense variants IMP_SYN | Synonymous IMP_INF | In-frame delins IMP_SPL | Splicing assessment IMP_FXN | Variant-specific Funct Assays ---- ### Point approach using Bayesian framework Weights will be given as points based on odds of pathogenicity ([Tavtigian et al. 2020](https://pubmed.ncbi.nlm.nih.gov/32720330/)). Conversion from SVC 3.0 codes to point equivalents is possible (allowing for migration), but will not be exact as point approach allows more granular selection. Comparison: SVC 3.0 strength | SVC 4.0 points ------ | ------ BS | -4 BP | -1 (ind.) | 0 PP | 1 PM | 2 PS | 4 PVS | 8 ---- ### Decision tree examples General structure: 1. Variant type 2. Assess predictive data (NMD, splicing, missense, ...) 3. Assess exon and transcript relevance (in all clinically relevant transcripts? MANE Select/Plus Clinical) 4. Assess region information (in critical domain/size) 5. Assess informative variants (other variants in same codon or similar impact) Examples: - [DRAFT-LoF-type-Decision-Trees](https://drive.google.com/file/d/1ymvftv7Pocvyd6KXNokFwxAukGlg93EP/view?usp=drive_link) - [DRAFT-Missense-Decision-Trees](https://drive.google.com/file/d/1KzywsJ15qNycLKPtIk-pVG_xIshQuBEB/view?usp=drive_link) - [Copy of ACMG_Forum_March2024](https://docs.google.com/presentation/d/1NpqUXHM_QLu0Xx3bDXJRtSeGS4omJrd4ZG4wiLUl2PU/edit#slide=id.g2619859af94_0_1487) - [AMP_2023_SVCUpdate](https://docs.google.com/presentation/d/1pEx7va4z1uDRoovPq0cbqhoR7HzTpuL4/edit#slide=id.p22) #### POP_FRQ Three different approaches to calculating Disease allele frequency (DAF): - Binning: Requires estimate of prevalence (similar to our current approach) - Prevalence: Requires estimates of prevalence, heterogeneity and penetrance - Pathogenic variants: Requires presence of known P/LP variants Then compare observed allele frequency to DAF: Result | Points ------ | ------ Consistent with AF of P/LP variants in this gene | POP_FRQ_0 \<3x but >= 1x of DAF | POP_FRQ_-1 \<10X but >= 3x of DAF | POP_FRQ_-3 \>= 10x of DAF | POP_FRQ_-6 **Significant difference from SVC 3.0**: PM2 (absence in healthy population) is translated to POP_FRQ_0, i.e. this line of evidence will no longer contribute any weight on its own. However, it will be a requirement for some of the other criteria. #### IMP_MSS and IMP_SPL IMP_MSS: 1. Evaluate presence in all/some relevant transcripts 2. Decide which predictor to use for a gene **prior to assessment** 3. Assign points according to translation table for predictor (see https://pubmed.ncbi.nlm.nih.gov/36413997/) 4. Compare with other, nearby variants IMP_SPL: 1. Evaluate if splicing impact is predicted 2. Evaluate NMD 3. Evaluate presence in all/some relevant transcripts (MANE Select/Plus Clinical) 4. Evaluate affected region 5. Compare with other, nearby variants NOTE: IMP_MSS and IMP_SPL are mutually exclusive - only one of them should be used for a given variant. Choose whichever gives the highest pathogenic score. Note also that both should be evaluated for missense variants. #### IMP_NUL/IMP_NMD 1. Evaluate pos (>-50 nt last exon-exon) 2. Evaluate alt start codon, other P LOF variants 3. Evaluate presence in all/some relevant transcripts 4. If NMD not predicted: Evaluate region (critical/functional domain) 5. Evaluate other P/LP variants with NMD in same exon ### Deciding clinical significance Clinical significance is simply decided based on the sum of points from applied criteria: Clinical significance | Point sum ------ | ------ Class 1 | <= -4 Class 2 | -3 to -1 Class 3 | 0-5 \> Low | 0-1 \> Mid | 2-3 \> High | 4-5 Class 4 | 6-9 Class 5 | >= 10 **Significant differences from SVC 3.0**: - -1 = Class 2, equivalent to 1 BP (previous -2 points = 2 BP) - Subcategorized Class 3: Not required, but will recommend labs to start tracking </details> ## Necessary changes ### Changes to ELLA's UI Nested choices for applying criteria, where each previous choice decides the available next choices. Some choices can most likely be automated. ### Changes to ACMG rules engine A complete rewrite will be necessary. Should take into account: - Mutually exclusive codes/criteria - Allowed point ranges/values for each criterion - Scenarios with/without available information from annotation ### Migration Existing codes can most likely be migrated to the new codes and points system without much effort, but needs validation.