Objective: We aimed to search for a possibility to modify Langbehn et al. model (LanM) for prediction of Huntington disease (HD) age of onset based on length of CAG repeat expansion using multidimensional optimization technique on the third Enroll-HD study periodic dataset (PDS3).

Background: Among a number of statistical models which establish relationship between CAG expansion length and HD age of onset, LanM appears to be the most accurate to date. However, it is applicable for those HD mutation carriers only with 36-56 CAG repeats, being the most accurate for 41-56 CAG repeats.

Methods: PDS3 database comprises the clinical data from 8,714 participants, including 3,598 from the European REGISTRY study that preceded Enroll-HD. We took data for all manifest HD mutation carriers with verified CAG expansions of at least 36 (n=4789). The median for CAG repeats was 43, ranging from 36 (min) to 71 (max). We used least squares method (LS) implemented to the SciPy library to determine the best curve fit based on data analyzed. The model has the following form: age = exp(x₀*CAG + x₁) + x₂. In this model, “age” is the date of clinical HD diagnosis, “CAG” is the number of CAG repeats, and “x” is the vector of model’s parameters. We took LanM’s coefficients for the initial guess on independent variables (-0.146; 9.556; 21.54, respectively). We applied mean squared error (MSE) as an average quality measure for all subjects and for 3 specified subgroups: subjects with 36-39 (subgroup A), 40-56 (subgroup B), and those who have 57 CAG repeats and more (subgroup C), respectively.

Results: After processing, we received the following modified model (ModM): age = exp(-0.084*cag + 7.292) + 9.406. In general, MSE for the LanM (MSE-L) was 67.75, whereas with application of ModM (MSE-mod) = 63.83. For the subgroups we obtained following results. Subgroup A: MSE-L = 447.98; MSE-mod = 285.57. Subgroup B: MSE-L = 60.20; MSE-mod = 59.20. Subgroup C: MSE-L = 38.91; MSE-mod = 28.20 (see [figure1]). However, on closer examination, we found that for carriers of 57-66 CAG repeats ModM is less accurate: MSE-L = 18.54, MSE-mod = 28.8.

Conclusions: The results suggest that the ModM is more accurate for 36-39 CAG repeats carriers in comparison with LanM. For 40-56 CAG repeats carriers, the ModM has similar accuracy to LanM. This allows to consider ModM as an alternative to LanM for prediction age of HD onset for 36-56 CAG repeats carriers.

« Back to 2017 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/a-modified-model-for-prediction-of-huntington-disease-age-of-onset-based-on-length-of-cag-repeat-expansion/