Summary
A medRxiv preprint combines single-molecule testing of 91 KIF1A variants with longitudinal clinical data from 343 patients to classify disease trajectories and predict adaptive-behavior scores.
A medRxiv preprint posted on September 16, 2026 reports a prognostic framework for KIF1A-associated neurological disorder (KAND), a rare monogenic neurological condition. The researchers combined laboratory measurements of 91 pathogenic KIF1A variants with longitudinal clinical characterization of 343 patients.
The resulting model classified patients into stable and declining clinical trajectories with an area under the receiver operating characteristic curve (AUC) of 0.834. It also predicted scores on the Vineland Adaptive Behavior Scales Adaptive Behavior Composite (VABS ABC), a measure of adaptive functioning, with R² = 0.484 under leave-one-out cross-validation.
How the framework was built
KAND is associated with disease-causing changes in KIF1A, which is involved in kinesin motor function. Kinesin proteins help transport material within cells, including neurons. The study used “single-molecule biophysical phenotyping” to generate quantitative functional measurements for individual pathogenic variants, then combined those measurements with computational pathogenicity scores and clinical variables.
The clinical analysis used the VABS Adaptive Behavior Composite and Growth Scales to examine how patients' outcomes changed over time. Two divergent trajectories emerged after age 10: one group remained relatively stable, while the other showed decline and was characterized by seizures, abnormal electroencephalograms (EEGs) and optic nerve atrophy.
The study therefore connected three kinds of information: the measured molecular behavior of a KIF1A variant, computational estimates of its pathogenicity and features observed in the patient. That combination is intended to provide more useful outcome estimates than a molecular diagnosis considered without clinical context.
What the model found and why it matters
The AUC of 0.834 describes the model's ability to distinguish the stable and declining groups in the reported analysis; an AUC of 0.5 corresponds to chance-level discrimination, while 1.0 represents perfect separation. For the continuous VABS ABC outcome, R² = 0.484 means that the model accounted for 48.4% of the variation in scores in the leave-one-out cross-validation analysis.
The authors present the framework as a possible prognostic tool for people with newly identified KIF1A variants. Prognostic information could also help researchers group participants by expected disease course when designing clinical trials, creating a more defined basis for comparing outcomes.
The proposed reach is broader than KAND. Because kinesin motor function is mechanistically conserved, the authors suggest that a similar strategy could be tested in other kinesinopathies and in monogenic disorders where quantitative functional assays are available. That extension is a research proposal from the preprint, while the reported performance comes from the KAND analysis.
This is an early research result reported as a medRxiv preprint. The model was evaluated with leave-one-out cross-validation in the described dataset, and the supplied abstract does not describe independent or prospective validation. How reliably the framework transfers to other patient populations or neurological disorders remains to be established.