ABOPM · ACMG

AI note — cites the primary sources.
The ClinGen Sequence Variant Interpretation (SVI) working group's published recalibration of PM5 (Abou Tayoun et al., Hum Mutat 2018) established the general framework but explicitly did not set a numeric threshold for PM5_strong escalation. The field has converged, not via a single normative document but via accumulated gene-specific expert panels, on ≥3 different pathogenic substitutions at the same residue as the typical threshold for PM5_strong. Gene-specific panels matter: BRCA1/2 VCEP allows PM5_supporting for some cases, PM5_strong only with 3+ substitutions AND at least one with ClinVar review status 2+ stars. For board preparation, memorize 3+ as the modal answer and be prepared to defend PM5_moderate as the conservative default when the evidence is not clearly residue-wide. — @CANONIC · governed AI layer · sources: Abou Tayoun et al. Hum Mutat 2018 (PM5 recalibration), ClinGen BRCA1/2 VCEP Criteria v1.1.0, Ellard et al. Eur J Hum Genet 2019 (implementation framework). Not a substitute for your institution's variant-curation SOP.

Marcus Freeman · Attending · Genomic pathology · PharmGKB CPIC contributor
@priya_chen this is a good question and the answer in practice is less consensus than the literature pretends. Our lab's internal SOP requires ≥3 different pathogenic missense changes at the same residue AND at least one of them with ClinVar 2+ star review status for PM5_strong; we hold at PM5_moderate otherwise. The ClinGen SVI working group working document suggests ≥3 but calls it a 'starting calibration' — my read is the field is standardizing around 3 but nobody has written the definitive implementation paper yet. For boards I'd memorize 3+ as the modal answer but be prepared to defend PM5_moderate as the conservative default. Prep hack: the 2019 Ellard framework paper in Eur J Hum Genet walks through the exact calibration logic and is a common exam source. #acmg_debate

Alexa Park · Fellow · Heme/Onc · breast focus
From the clinical side — when our molecular pathology colleagues escalate to PM5_strong, that often flips our treatment-decision certainty from 'consider germline testing' to 'order germline testing now.' The calibration matters downstream. +1 to @marcus_freeman's 3-different-residue-changes rule — that's what I've seen at our NCI-designated tumor board as the de facto standard.

AI note — cites the primary sources.
ClinGen TP53 VCEP Specifications v1.4.0 (2023); Giacomelli AO et al Nat Genet 2018; Kato S et al PNAS 2003. Residents should note: the ClinGen VCEP framework is now the authoritative source for TP53 germline classification, superseding older ACMG-generic approaches.

AI note — cites the primary sources.
ClinGen InSiGHT MMR VCEP 2019; Jaganathan K et al Cell 2019 (SpliceAI). Splice-proximal variants (+5 position) warrant extra caution — SpliceAI has lower specificity for +5 than canonical +1/+2. Consider a minigene assay if RNA not available.

Marcus Freeman · Attending · Genomic pathology · PharmGKB CPIC contributor
SpliceAI 0.34 is below the typical 0.5 threshold for confident splicing prediction. Without RNA data, this stays VUS — but segregation in 3+ affecteds can pull PP1_strong, pushing to LP. Get the segregation data before the RNA assay. Reference: ClinGen InSiGHT MMR VCEP classification criteria 2019.

AI note — cites the primary sources.
Richards S et al Genet Med 2015 (ACMG/AMP); Abou Tayoun AN et al Hum Mutat 2018 (PVS1 refinement). The 2018 PVS1 refinement explicitly addresses frameshifts distal to the last 50bp of the penultimate exon — invoke PVS1 strong not moderate when NMD-predicted.

Alexa Park · Fellow · Heme/Onc · breast focus
PVS1 applies (frameshift in canonical transcript, no predicted rescue isoform) + PM2_supporting (gnomAD absent). That gets you LP per ACMG/AMP 2015 rules (2 moderate-strong lines of evidence). You do not need functional data for a clean PVS1 call. Reference: Richards S et al Genet Med 2015;17:405.