Background Machine learning (ML) enables the discovery of patterns and relationships within large-scale datasets and can turn information into valuable actionable knowledge. In biomedicine, ML has proven beneficial for the prognosis and diagnosis of different diseases, including cancer and neurodegenerative disorders. Enroll-HD has become a large, high quality dataset, describing over 20.000 participants. As such, it is amenable to ML methods and could be used to produce new insights into disease progression that could have a high impact on HD patients and their carers.

Aim We analyzed the PDS5 in a large scale manner using ML algorithms to obtain valuable insights to the manifestations of HD for personalized treatments and prognoses, for manifest and premanifest patients.

Methods We applied ML methods for data preprocessing and for making predictions. We also applied recurrent neural networks to demonstrate how neural networks can learn from temporal data to enrich future predictions.

Results The application of ML methods for data pre-processing can accurately impute the missing values. We further explored the utility of ML to predict Age of Onset in comparison to the widely implemented Langbehn formula and for making personalized predictions regarding the driving capability of HD patients. The resulting pre-processed Enroll-HD dataset as well as the pre-processing workflow are available to be used for related HD-disease predictions.

Conclusions ML can assist in making predictions and increasing the quality of a dataset to be ‘ML-ready’. Our predictions can be used to implement an advisory system that can assist both clinicians and patients in decision making.