Clinical actionability — for cohorts and clinicians
Why this matters for PACS1 patients today
Many PACS1 patients are already on phenytoin, clonazepam, or oxazepam — empirically. The clinical question is: do these patients have meaningfully different outcomes (in seizure control, quality of life, side effects) than patients on other AEDs?
A retrospective cohort analysis can answer this question at zero cost in weeks. The data are already collected in specialized PACS1 clinics. The comparison structure:
Treatment | Predicted seizure control | Predicted behavior |
Phenytoin | Better than expected for AED noise floor | — |
Clonazepam | Better than expected | + anxiolytic / behavioral |
Oxazepam | (Variable, off-label for epilepsy) | + anxiolytic / behavioral |
Carbamazepine | Standard | — |
Lamotrigine | Standard | — |
Levetiracetam | Standard | — |
Valproate | Standard | — |
Diazepam, lorazepam, midazolam | Standard | — |
Other antiepileptics | Standard | — |
The prediction's testability comes from the narrowness of the predicted-positive set (3 drugs of 19 AEDs tested), the carbamazepine negative prediction (often used in PACS1 — these patients should NOT differ from baseline), and the clonazepam vs diazepam contrast (two benzodiazepines; only clonazepam predicted selective).
Practical proposal for collaborating clinics
A typical analysis structure:
- Identify all PACS1 patients in the cohort with documented R203W mutation
- Stratify by AED treatment (phenytoin / clonazepam / oxazepam / carbamazepine / other)
- Compare outcomes: seizure frequency reduction, quality-of-life metrics, side effects, treatment retention
- Adjust for confounders (age, severity, polytherapy, treatment duration)
- Publish results in pediatric neurology or rare-disease journals
This is realistic for clinics with > 20 PACS1 patients and matched outcome data. The Schuurs-Hoeijmakers consortium (NL) and CHU Montpellier (FR) are obvious candidates.
What if the prediction holds
If the retrospective cohort shows better seizure outcomes in phenytoin/clonazepam/oxazepam patients, it would be the first mechanistically-grounded therapeutic recommendation for PACS1. Guidelines could shift toward these drugs as preferred first-line. Validation experiments (SPR binding, cellular rescue) become high-priority and well-funded. The framework can be extended to other rare missense diseases.
What if the prediction fails
If the cohort shows no difference, the in silico hypothesis is wrong (or the wet-lab pocket is wrong, or the population is too small). The reliability scores need downgrading. The project pivots toward direct binding assays to test what went wrong.
This is a falsifiable prediction. Both outcomes are scientifically informative.