- Fold / Class
- Soluble (p=1.00), non-enzyme (p=0.99); part-ordered, part-disordered (~29% disorder). Its AlphaFold model is low-confidence (mean pLDDT ≈49).
- Tractable Pocket
- High-confidence pocket (0.80) at 122–136. A concrete, foldable handle on a protein AlphaFold barely resolves — distinct from the known N-terminal Aβ-binding region.
- Biology
- "Aggregatin": binds Aβ via an N-terminal domain (res 1–80) and nucleates Aβ aggregation; knockdown reduces plaques and cognitive deficits in AD models.
- Structure Map
- The order/disorder split is the construct-design plan for a protein with no experimental structure.
- Clean Signal
- Soluble; no membrane signal.
Model-reported confidence for the headline calls (amber = the load-bearing prediction the rest of the profile builds on). These are model-estimated probabilities that rank and gate each call — not calibrated rates of experimental success.
The Gap
Why This Target Is Still Dark
FAM222A — "Aggregatin" — is one of the more mechanistically interesting dark proteins in Alzheimer's disease. It accumulates in amyloid plaques, binds Aβ through an N-terminal domain (residues 1–80), and nucleates Aβ aggregation; its knockdown reduces amyloid deposition, neuroinflammation and cognitive deficits in AD models, and a variant in the gene associates with brain atrophy. Yet FAM222A is otherwise nearly uncharacterised (near-empty UniProt annotation; its sister gene FAM222B has no assigned function), sits at IDG Tdark, and has no experimental structure in the PDB. Its AlphaFold model is low-confidence (mean pLDDT ≈49) — so even the predicted structure is, in practice, a blank. A published effort had to build an ab-initio model to attempt virtual screening.
That combination — a validated Aβ therapeutic target that no model resolves — is exactly where prediction earns its keep: everything below is computed from the canonical 452-residue sequence and derived structural predictions, with no experimental FAM222A structure used as input.
Architecture & Topology
How the Sequence Is Organised
| Element | Residues | Note |
|---|---|---|
| Aβ-binding domain | 1–80 | N-terminal domain that binds Aβ and nucleates aggregation; mediates the plaque biology. |
| Ordered pocket region | 122–136 | High-confidence pocket; ordered and targetable, distinct from the Aβ-binding domain. |
| Remainder | ~137–452 | Largely disordered or modelled at low AlphaFold confidence — the natural regions to truncate for a well-behaved construct. |
The Predicted Pocket
The Predicted Pocket
A high-confidence pocket (0.80) that matters for two reasons. First, it is ordered in a protein that is largely not, so it is the natural focus for expression, biophysics and any small-molecule effort. Second, it is not the Aβ-binding region, so it offers an independent site to probe FAM222A's function or to modulate it without directly competing with Aβ. The one high-confidence pocket in a protein AlphaFold models at low confidence; no known small-molecule binder.
Site: Residues 122–136 (ordered region, distinct from the Aβ-binding domain)
Post-Translational & Structural Features
Specific, Testable Residues
- Mixed order/disorder (~29%). The disordered stretches are the natural truncation points; the ordered core (including the 122–136 pocket) is the expressible unit.
- No membrane signal. A soluble, cytosolic/nuclear protein.
- An independent, targetable pocket. Separate from the Aβ-binding domain — useful for probing function without competing with Aβ binding directly.
Recommended Experimental Follow-Up
An Orphan Sequence, Turned Into a Ranked Plan
Each prediction is paired with the experiment that would test it and the readout to watch for.
| Prediction | Experiment | Readout |
|---|---|---|
| Ordered core + pocket (122–136) | Express the ordered core; biophysics / fragment screen | A well-behaved construct + a hit at the pocket |
| ~29% disorder map | Limited proteolysis / truncation series | Confirmed foldable boundaries |
| Pocket is separate from the Aβ domain | Pocket-mutant vs Aβ-binding assay | Function independent of Aβ binding |
| Aβ-nucleation via res 1–80 | N-terminal-domain constructs ± Aβ | Confirm the nucleation interface |
Scope & Limitations
What This Is — and Isn't
- Prediction, not experiment. These are computational hypotheses to prioritise experiments — not a substitute for a structure or an assay. No result here has been validated in the wet lab.
- The pocket is predicted; no binder is named. The 122–136 site is a residue-level hypothesis for a foldable, targetable region — not a demonstrated ligandable pocket.
- Biology caveats. The in-vivo Aβ / plaque data are from knockdown (AAV-shRNA), not a germline knockout; the therapeutic case is a hypothesis.
All predictions were generated with Orbion's Astra suite from the canonical FAM222A sequence (UniProt Q5U5X8), using AlphaFold-derived structural features. Reported values are model outputs; model internals are out of scope.
References
- [1]UniProt Consortium. UniProtKB entry Q5U5X8 (FAM222A, human). uniprot.org.
- [2]Pharos / Illuminating the Druggable Genome. FAM222A target record — Tdark. pharos.nih.gov.
- [3]Yan T., Liang J., Gao J., et al. FAM222A encodes a protein which accumulates in plaques in Alzheimer's disease. Nat. Commun. 11, 411 (2020). https://doi.org/10.1038/s41467-019-13962-0
- [4]Alabdulraheem Z.T.J., Durdagi S. Ab initio and comparative 3D modeling of FAM222A-encoded protein and target-driven virtual screening for Alzheimer's disease. J. Mol. Graph. Model. 125, 108575 (2023). https://doi.org/10.1016/j.jmgm.2023.108575