Skip Navigation

ALS Alert Newsletter

field with sun shining

Jun 7

Answer ALS annual symposium highlights promise and excitement

Annual meeting held virtually and welcomed over 100 researchers.

Although the meeting was virtual, the excitement at the Answer ALS annual symposium was very much real. Even a global pandemic didn’t slow down the scientists working to move ALS science forward. The biggest announcement was that the data is here. On January 8, 2021, the Answer ALS data portal officially launched, and researchers around the world are already hard at work using the information.

Scientists are also making advances in other areas, too. Sophisticated artificial intelligence and machine learning analyses of Answer ALS data has enabled researchers to create more in-depth phenotypes of ALS data and identify genetic variants with small contributions to disease that might act in tandem with multiple other variants. In short, ALS hasn’t been solved, but progress over the past year has brought us closer than ever.

Answer ALS is governed by the same ethos as the Packard Center, which emphasizes openness and collaboration as key steps to move the science forward. To encourage researchers to share their findings, even if the results aren’t final or they reveal challenges, the Answer ALS meeting is kept closed to the general public. This means that specifics of some of the conclusions are kept under wraps until they have been vetted by other scientists as part of peer-review and published in a research journal.

One of the challenging parts of studying ALS and understanding the disease is its underlying heterogeneity. Less than 15% of patients have a known genetic cause of ALS, and there can be wide variability in disease progression. Despite this, genetics studies have shown that sporadic ALS is between 40 and 60% heritable. This indicates the possibility for underlying genetic and molecular patterns that can be used to cluster patients. Scientists have begun to try and draw connections between genotype and phenotype, a task that is only possible with the kind of large and detailed datasets provided by Answer ALS.

These kinds of subtle connections between multiple genetic variants require machine learning and other advanced computational methods to try and understand the complexities hidden in the vast amounts of patient data. One of these priorities is finding ways to model disease progression in different groups of patients, which will be key to testing therapeutics. Existing models, however, don’t provide the needed precision. With Answer ALS data, researchers have begun identifying different subgroups of patients to create disease progression clusters with varying patterns of decline based on ALSFRS-R scores and other clinical data. Preliminary data analysis has identified several interesting non-linear patterns of decline that correspond to alternative disease progression metrics such as survival outcomes.

Machine learning techniques are also being used to detect small contributions from a wide range of genetic variants. Traditional Genome-Wide Association Studies (GWAS) can measure large effects from a small number of genetic variants, which means that scientists don’t have answers about the underlying cause of disease in large numbers of patients. Even advanced strategies like rare variant burden analysis, which adds the cumulative effect of multiple rare variants into a single genetic score, doesn’t capture the complexity of multiple genetic variations. But by training a machine learning model on the genetic features of individuals with ALS, scientists can look at the impact of multiple genetic variants at once. In an iterative process, the researchers have created a pipeline called Pathways-based Recursive Feature Selection (PathRFS) to take the top genetic variants from the top pathways to train new models. When scientists used PathRFS on rare exonic variants to classify ALS cases vs. controls, they found they could moderately predict ALS clinical progression and extreme ALS phenotypes.

Scientists have also been comparing ALS cases with controls at the RNA and protein level, characterizing early RNA differences that persist to the end of disease. These same machine learning-based comparisons on ALS iPSC lines have also yielded results. By looking at physical features of cells, researchers are working to identify an ALS signature in patient-derived motor neurons. The existing model can discriminate between cases and controls with moderate accuracy. Scientists hope that combining multidimensional feature analysis with other ’omics data could lead to leaps and strides in patient stratification.

All of these data, however, won’t be helpful if scientists can’t access it. It’s why one of the major goals of Answer ALS is to create a data portal that integrates resources and creates a collaborative, cloud-based workspace.

In 2020, the Answer ALS data portal team focused on making it easier to find and access Answer ALS data. Forming the core of the portal is the data from the 1041 Answer ALS participants, which includes ’omics data and iPS cell line information. The new portal contains new search and filter functions, and a simpler download and data transfer process in mind. Importantly, the system was also designed with expansion in mind, and the team is already working on that development. In the coming year, the Answer ALS team is focused on expanding and improving this industrial-strength data portal to create a centralized ALS data hub. Forthcoming improvements include integrating tool ordering with the portal, and adding more samples from outside Answer ALS to create a one-stop shop for ALS research.

The whole Answer ALS team is looking forward to gathering in person in Baltimore next year.