Machine Learning Helps Identify Patterns of Progression in ALS

According to a story from, neurodegenerative diseases, such as amyotrophic lateral sclerosis, are diseases that progressively worsen over time. However, progression in this disease is not consistent across all cases—some patients progress at different speeds or in different ways. A better understanding of progression in amyotrophic lateral sclerosis and other diseases similar to it, like Parkinson’s or Alzheimer’s, is important for understanding root causes, enrolling patients in clinical trials, and evaluating possible treatment approaches. In a recent study, a research team developed a machine learning platform that can help scientists understand patterns of progression in these types of diseases.

About Amyotrophic Lateral Sclerosis (ALS)

Amyotrophic lateral sclerosis, otherwise known as Lou Gehrig’s disease, is a rare, degenerative disease that causes the death of nerve cells associated with the voluntary muscles. Little is known about the origins of amyotrophic lateral sclerosis, with no definitive cause in about 95 percent of cases. The remaining five percent appear to inherit the disease from their parents. Symptoms initially include loss of coordination, muscle weakness and atrophy, muscle stiffness and cramping, and trouble speaking, breathing, or swallowing. These symptoms worsen steadily over time; most patients die because of respiratory complications. Treatment is mostly symptomatic and the medication riluzole can prolong life. Life expectancy after diagnosis ranges from two to four years, but some patients can survive for substantially longer. To learn more about amyotrophic lateral sclerosis, click here.

About the Research

Many clinical trials happening today assume a linear rate of progression in neurodegenerative disease, but the machine learning technique found that this wasn’t the case. The scientists were able to develop ALS disease progression subtypes that were consistent across different disease metrics and patient populations. The team applied the model to five different patient data sets from observational studies and clinical trials. 

The researchers named the four different progression subtypes:

  • Stable slow progression
  • Unstable slow progression
  • Unstable moderate progression
  • Sigmoid fast progression

Many of them had a distinctly nonlinear character. Some of these trajectories are characterized by a very sudden loss of function that has major impacts on quality of life, ability to participate in trials, and treatment methods.

The team was also able to effectively use the machine learning approach with Parkinson’s and Alzheimer’s as well; in Alzheimer’s, the scientists found varied rates of transition from what was considered mild disease to severe disease. When used on Parkinson’s, the team identified disease phenotypes and off-medication scores as correlating with trajectories of progression.

The machine learning method used in this research gives scientists new ways of understanding neurodegenerative illness that could help in identifying subtypes and improve clinical trial effectiveness.

Share this post

Share on facebook
Share on twitter
Share on linkedin
Share on pinterest
Share on print
Share on email