Machine Learning Data Analysis for Parkinson’s Prediction

Parkinson’s progression – let’s put the data to use

Parkinson’s disease progression has long puzzled the medical community. With an influx of data, papers and researchers our team decided to undertake the task of getting closer to discovering the nature of this disease by analysing biomarker data indicative of the disease’s advancement.

Goal – recognise Parkinson’s before it strikes

The challenge lies in identifying novel biomarkers contributing to Parkinson’s progression. It’s believed that variations in protein and peptide levels are crucial indicators of the disease’s development. The aim is to predict future MDS-UPDRS (Unified Parkinson’s Disease Rating Scale) scores for patients, providing a roadmap for devising effective, personalised treatments, given the absence of a Parkinson’s cure.


Solution – Machine Learning model that predicts disease

Despite existing monitoring methods, predicting Parkinson’s progression with certainty remains elusive. 

Our specialist proposed a Machine Learning (ML) model capable of predicting individualised MDS-UPDRS scores based on patients’ protein and peptide levels. 

To make it happen we’ve analysed data from over 10,000 subjects, including patients’ peptides/proteins levels (taken from Cerebrospinal Fluid samples) and past UPDRS scores.

This required meticulous data analysis to reveal hidden patterns and correlations, which were instrumental in training ML algorithms for predicting Parkinson’s progression at specified intervals.

Python

Scripting

Pandas

Data handling

CatBoost

Decision tree building

XGBoost

Decision tree building

Optuna

Hyperparameter optimisation

TabNet

Tabular data pattern identification

Results – accurate prediction aiding diagnosis

A functional ML model was developed, capable of predicting Parkinson’s progression. Starting with raw data, the team imputed missing records, trained algorithms, and successfully made accurate predictions about Parkinson’s progression, providing a valuable tool for diagnosis and treatment improvement.

This endeavour was not only an important learning experience but also an application of data analysis and ML knowledge to a real-world challenge, making a positive impact on individuals with Parkinson’s disease.

For a deeper understanding and collaboration on similar projects, please get in touch


Or contact us right now and let’s build your product together!

Your inquiry