# NM_001368809.2(AMPD2):c.1858C>A (p.Arg620Ser)

https://sugi.bio/variant/ampd2-p-arg620ser

## Summary

**AMPD2 p.Arg620Ser (c.1858C>A)** is classified **Likely pathogenic** in AMPD2 (ClinVar, Single submitter, 1 submitter; last evaluated 2023-05-27). AlphaMissense: likely pathogenic (0.9981); Absent from gnomAD v4.1; Associated with pontocerebellar hypoplasia type 9.

**At a glance:** Likely pathogenic · review: Single submitter · rsID rs756745194 · 1 submitter · 1 linked condition · AlphaMissense likely pathogenic · absent from gnomAD v4.1

## For patients & families {#patients}

This is a change in the AMPD2 gene that is considered likely disease-causing, submitted by 1 submitter, and is linked to pontocerebellar hypoplasia type 9.

*Reference information, not medical advice, and not a prediction for any individual. A genetic counselor or clinician can interpret what it means for you and your family.*

**Support & next steps:** [Condition overview & support (NIH GARD)](https://rarediseases.info.nih.gov/diseases/0017590/index) · [Find a genetic counselor (NSGC)](https://findageneticcounselor.nsgc.org/).

## Identity {#identity}

| Field | Value |
| --- | --- |
| Gene | [AMPD2](https://sugi.bio/atlas/gene/AMPD2/) |
| Protein change (HGVS p.) | p.Arg620Ser |
| Coding change (HGVS c.) | c.1858C>A |
| MANE Select transcript | [NM_001368809](https://www.ncbi.nlm.nih.gov/nuccore/NM_001368809) |
| dbSNP | [rs756745194](https://www.ncbi.nlm.nih.gov/snp/rs756745194/) |
| Variant type | single nucleotide variant |
| Location | chr1:109629486-109629486 (GRCh38) |
| ClinVar | [VCV2503047](https://www.ncbi.nlm.nih.gov/clinvar/variation/2503047/) |

**All HGVS expressions:** `NC_000001.10:g.110172108C>A`, `NC_000001.11:g.109629486C>A`, `NG_034075.1:g.14674C>A`, `NG_088585.1:g.262C>A`, `NM_001257360.2:c.2020C>A`, `NM_001257361.2:c.1666C>A`, `NM_001308170.1:c.1795C>A`, `NM_001368809.2:c.1858C>A`, `NM_004037.9:c.1858C>A`, `NM_139156.4:c.1777C>A`

## Computational & population evidence {#evidence}

**Concordance:** 1 independent predictor **concordant** with the ClinVar classification.

- AlphaMissense **likely pathogenic** (0.9981) — concordant with the pathogenic call
- REVEL 0.897 (moderately pathogenic-leaning) — agrees with AlphaMissense
- → **In-silico consensus:** AlphaMissense and REVEL both call this damaging; SaProt makes no call
- Conservation: phyloP 7.686, GERP 0.065, phastCons 1 — highly conserved (agrees)
- **SaProt:** not scored — SaProt is structure-aware and AlphaFold DB has no model for this protein. That is a property of the dataset, not of the variant, and is not evidence either way.
- ⚠ **Isoform check:** AlphaMissense is numbered on a different transcript (AM **R674S** vs ClinVar **R620S**) — the AlphaMissense read here may be for a different residue; verify against the ClinVar transcript.
- AlphaMissense ranks this among the **top 4%** most-pathogenic-predicted substitutions in AMPD2 (of 5,188 modeled).
- CADD: **PHRED 29.3**. No single ClinGen-calibrated threshold exists and the expert panels that use CADD disagree — GUCY2D/AIPL1 require ≥ 20.0 for an untranslated-region variant, ABCA4 ≥ 28.1 for a synonymous or indel variant, the Pulmonary Hypertension panel imports Pejaver 2022's missense-calibrated numbers, and the Antibody Deficiencies panel declines CADD entirely for PIK3CD. CADD v1.7 also incorporates Zoonomia phyloP, so it partly re-expresses the conservation above. *Descriptive; no threshold applied.*
- Regional missense constraint: **o/e 0.71** across the covering region of the MANE Select transcript (109,628,140–109,631,149), spanning 3,009 bp. Chi-square p 8.7e-13. Describes the region, not this allele; gnomAD v4.1.1's own constrained determination. *Descriptive; no ACMG code applied.*
- SpliceAI: **no splice effect predicted**. SpliceAI outputs four numbers per variant, not one: acceptor loss, donor loss, acceptor gain, donor gain (DS_AL / DS_DL / DS_AG / DS_DG). A loss score is an existing splice site weakened or destroyed; a gain score is a new site created or an existing one strengthened. The release publishes its own filter rule, which we read from the dataset rather than assume: an allele is reported only if the largest of its four deltas reaches 0.2. No allele at this position is in the release, so all four are under 0.2 for every alternate allele here. No figure is shown because none was read: the bound is deduced from that absence. It is a result, not a gap in our data, because the release scores every possible substitution across the transcripts it covers (Ensembl MANE v1.4) before it filters, so what survives for a gene is one continuous stretch. What is measured is that this position sits inside the stretch covered for AMPD2 (873 scored positions, the nearest 2 bases away); what is taken from the release's own description, rather than measured, is that coverage inside that stretch is continuous. On that, SpliceAI did score it. *ClinGen's SVI treats the largest of the four as evidence against a splicing effect at or below 0.1, where it may support BP4, and as uninformative between 0.1 and 0.2. We cannot tell which applies here, so these SpliceAI data support neither BP4 nor PP3; that is a statement about this evidence alone, not about the variant's other evidence.*
- Population frequency: **absent from gnomAD v4.1** — a rarity signal (ACMG PM2-supporting), not evidence of pathogenicity on its own.

*Computational predictors are not independent (ClinGen Variant Classification guidance): AlphaMissense (Cheng et al. 2023) carries the ACMG weight for missense and conservation (phyloP/GERP/phastCons) for non-missense; REVEL (Pejaver-2022 calibrated strength bands) and SaProt (Su et al. 2023, structure-aware, ClinVar-independent protein language model, raw LLR) are shown as orthogonal agreement signals, not additive. For frequency, the gnomAD v4.1 grpmax filtering allele frequency (FAF95) is ClinGen's designated BA1/BS1 input (gnomAD v4 guidance v3.0, June 2025); BA1/BS1 thresholds are disease-specific, not a blanket 5%. Predictions, not a clinical determination.*

## Gene constraint & dosage {#gene-context}

**AMPD2** population constraint: LOEUF 0.705, pLI 1.3901e-08, missense-Z 3.6393. *Lower LOEUF and higher pLI indicate a gene that tolerates loss-of-function poorly; ClinGen dosage is the curated haploinsufficiency call.*

## Protein context {#protein}


AlphaFold model confidence (whole protein): pLDDT 81.28, 59% of residues very-high.

**5 experimental structure(s)** (PDB): [26XC](https://www.rcsb.org/structure/26XC), [4NO3](https://www.rcsb.org/structure/4NO3), [4NO5](https://www.rcsb.org/structure/4NO5), [8HU6](https://www.rcsb.org/structure/8HU6), [8HUB](https://www.rcsb.org/structure/8HUB).

## What this gene does {#gene-function}

*Describes the biological roles of AMPD2 (the gene this variant disrupts) — not effects measured for this specific variant. A damaging variant is expected to affect these functions and pathways; the degree depends on the variant.*

**How this variant is thought to act:** AMPD2's established molecular role includes AMP deaminase activity.


### Pathways affected {#pathways}

AMPD2 participates in **1 Reactome pathway** (⚕ = disease pathway):

| Pathway (Reactome) | Evidence |
| --- | --- |
| [Purine salvage](https://reactome.org/content/detail/R-HSA-74217) | curated |

### Molecular function & processes (GO — experimentally supported) {#go}

**Molecular function:** AMP deaminase activity, identical protein binding.

**Biological process:** IMP biosynthetic process, GMP salvage, cyclic purine nucleotide metabolic process, energy homeostasis.

## Clinical significance {#significance}

**Likely pathogenic** — review status Single submitter *(criteria provided, single submitter)*, last evaluated 2023-05-27.

**Submitter consensus:** unanimous (1/1 Likely pathogenic).

### Submitter classifications {#submitters}

| Submitter | Classification | Review | Method | Date |
| --- | --- | --- | --- | --- |
| Center for Comprehensive Genetic Services, Shahid Beheshti University of Medical Sciences | Likely pathogenic | criteria provided, single submitter | clinical testing | 2023-05-27 |

## Associated conditions {#conditions}

[pontocerebellar hypoplasia type 9](https://sugi.bio/atlas/disease/pontocerebellar-hypoplasia-type-9/).

**Reported phenotype names (ClinVar):** Pontocerebellar hypoplasia type 9.

## Same-residue context {#hotspot}

Residue **620** carries **2 other pathogenic ClinVar variants** — a recurrently-mutated position: [AMPD2 p.Arg620Leu](/variant/ampd2-p-arg620leu), [AMPD2 p.Arg620His](/variant/ampd2-p-arg620his). *Positional co-occurrence of independent ClinVar records, not functional proof.*

## Similar variants {#similar}

Related AMPD2 variant pages (same residue, condition, or type): [AMPD2 p.Arg620His](/variant/ampd2-p-arg620his), [AMPD2 p.Arg620Leu](/variant/ampd2-p-arg620leu), [AMPD2 c.1698+1G>A](/variant/ampd2-c-1698-plus-1g-a), [AMPD2 p.Arg353Ter](/variant/ampd2-p-arg353ter), [AMPD2 p.Arg449Ter](/variant/ampd2-p-arg449ter), [AMPD2 p.Arg789His](/variant/ampd2-p-arg789his), [AMPD2 p.Glu174Ter](/variant/ampd2-p-glu174ter), [AMPD2 c.1080+3G>A](/variant/ampd2-c-1080-plus-3g-a).

### Data sources & attribution {#sources}

ClinVar (NCBI — public domain, built 2026-10-14); gnomAD v4.1 (Broad Institute, built 2026-10-15); AlphaMissense (Cheng et al. 2023, © Google DeepMind — CC BY 4.0, built 2026-01-27); REVEL (Ioannidis et al. 2016 — ODbL, built 2026-07-13); SaProt (Su et al. 2023 — MIT, built 2026-10-14); SpliceAI (Illumina — free for non-commercial use, built 2026-10-13); CADD (Rentzsch et al. 2019, University of Washington — free for non-commercial use, built 2026-10-10); Conservation (phyloP/phastCons/GERP) (UCSC hg38 470-way mammal (Hiller Lab) + Ensembl Compara, built 2026-07-12); gnomAD Non-coding Constraint (Gnocchi) (Chen et al. 2024, Broad Institute — CC0, built 2026-10-11); gnomAD Regional Missense Constraint (Broad Institute — CC0, built 2026-10-13); Pangolin (Zeng & Li 2022 — CC BY 4.0, built 2026-10-11); Repeat regions (UCSC — RepeatMasker, TRF, segmental duplications, built 2026-10-11); ClinGen (ClinGen, built 2026-09-22); UniProt (CC BY 4.0); Reactome / GO (CC BY 4.0, built 2026-09-23); MaveDB (per-record license, built 2026-10-11); CIViC (CC0, built 2026-09-22); Orphanet (Orphanet, built 2026-06-12)

*Primary record: NCBI ClinVar (variation 2503047). Classifications reflect these databases as of the page's build date and may change. Free reference use — not medical advice; consult the primary submitters and a clinician.*