RandomForest4Life: a Random Forest for predicting ALS disease progression.

Area of Science:

• Neurology
• Biostatistics
• Computational Biology

Background:

• Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease.
• Accurate prediction of ALS progression is crucial for patient management and clinical trials.
• Existing prediction models often lack sufficient accuracy and robustness.

Purpose of the Study:

• To develop and validate a predictive model for ALS disease progression.
• To identify key predictors of future disease advancement in ALS patients.
• To evaluate the model's performance in an external validation setting.

Main Methods:

• Utilized a random forest algorithm for prediction.
• Employed longitudinal patient examination data over three months.
• Trained and validated the model on data from 1197 ALS patients (PRO-ACT database).
• Externally validated the model using data from 625 additional patients.

Main Results:

• The developed method achieved third-best prediction accuracy in a challenge setting.
• Past disease progression was confirmed as a significant predictor of future ALSFRS scores.
• Greater variability in initial ALSFRS scores correlated with faster future disease progression.

Conclusions:

• Random forest models can effectively predict ALS disease progression.
• Multidimensional analysis of ALSFRS scores may enhance prediction accuracy.
• The findings support the use of past progression and score variability as key predictive factors.