Advanced learning models using Patient profiles and disease progression patterns for prognostic prediction in ALS



Amyotrophic Lateral Sclerosis (ALS) is a rapidly progressive neurodegenerative disorder characterized by the progressive loss of motor neurons in the brain and spinal cord, leading to muscular weakness and ultimately death. Life expectancy is 3 to 5 years after disease onset, there is no cure and its cause is uncertain. Disease heterogeneity makes it difficult to understand its underlying mechanisms. ALS is also the most common neurodegenerative disease in young adults. Thus, finding a cure/ways to slow disease progression, and promote patients’ quality of life are nowadays research challenges. In this scenario, rapid advances in computer/data science, genomics, imaging, and other technologies brought the promise of precision medicine (PM) in ALS.

PM, “an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person”, aims to predict more accurately which treatment and prevention strategies work in which groups of people. In this context, time-sensitive analysis of heterogeneous sources of genotype-phenotype data, clinical temporal data and remote patient monitoring data is recognized as the next critical step to enable the development of automatic prognostic methods based on patient profiles and disease progression patterns. Since the progression rate of ALS is highly variable, with patients surviving less than 6 months and others more than 30 years, biomarkers and disease patterns associated with disease heterogeneity are poorly characterized, impairing clinical trials’ results. Known prognostic factors related to shorter survival include bulbar-onset, shorter disease duration at diagnosis, older age at onset, early respiratory involvement, and faster functional decline. Nevertheless, despite enormous advances made recently in understanding ALS, the information about clinical and biomarker differences between patients grouped by disparate rates of disease progression is incomplete, and knowledge on how treatments should be optimised to manage patients with different progression rates in order to provide the best-suited treatment for individual patients according to a PM approach still does not exist.

In this scenario, AIpALS aims to advance PM and improve supportive care in ALS by proposing advanced machine learning (ML) models using patient profiles and disease progression patterns for prognostic prediction in ALS. These models will be learned from a large dataset of genotype-phenotype data and clinical temporal data already collected by national FCT project NEUROCLINOMICS2 (2016-03/2020, PI Sara C. Madeira) and European JPND project OnWebDuals (2016-2019, PI Mamede de Carvalho), together with new clinical data collected at ALS clinic (Centro Académico de Medicina de Lisboa) during this project, and new data gathered from remote patient monitoring using TeleHealth techniques. The knowledge gathered from learning patient profiles and disease progression patterns, together with the emergent prognostic prediction models will be made available to ALS clinicians in AIpALS App – a Telemedicine App for Decision Support and Knowledge Discovery in ALS. This integrative and interactive DS-KD system, continuously updated with data from patients’ follow-up at ALS clinic, and further integrating data from remote patient monitoring, will enable knowledge discovery using state-of-the-art data analysis and visualization tools, and prognostic prediction based on advanced ML, thus targeting PM and patient supportive care in ALS at the national level in an unprecedented way.

The scientific contributions of AIpALS have strong implications for both data science and clinical research communities, as well as measurable short-term impact as the resulting DSKS system can be made available to support computer-aided clinical decisions in national hospital settings. The expected impact is three-fold: computational (new efficient and effective integrative methods to identify patient groups from genotype-phenotype data and mine temporal patterns from clinical temporal data and remote patient monitoring data, together with advanced prognostic prediction models), social (better prognostic predictions to aid in clinical decisions and supportive care), and economic (better scheduling of medical appointments, personalised therapeutics and supportive care based on patients’ profiles and disease progression patterns). The team reflects the highly multidisciplinary of AIpALS, gathering unique computational expertise in data science/ML applied to neurodegenerative diseases (FCiências.ID – Data and Systems Intelligence and Health and Biomedical + Informatics Research Lines); Telemedicine/Decentralised Health Technology (FhP-AICOS – Assistive Information and Communication Solutions); and key clinical expertise (IMM team – Translational Clinical Physiology Lab with renowned ALS researcher).



Associação Fraunhofer Portugal Research (Fraunhofer), Instituto de Medicina Molecular João lobo Antunes (IMM/FM/ULisboa)


Funded by:


For any additional information, please contact us using the inquiries form.