Identifying Therapeutic Targets for Amyotrophic Lateral Sclerosis Through Modeling of Multi-Omics Data

Abstract

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease that primarily affects motor neurons, leading to loss of muscle control, and, ultimately, respiratory failure and death. Despite some advances in recent years, the underlying genetic and molecular mechanisms of ALS remain largely elusive. In this respect, a better understanding of these mechanisms is needed to identify new and biologically relevant therapeutic targets that could be developed into treatments that are truly disease-modifying, in that they address the underlying causes rather than the symptoms of ALS. In this study, we used two approaches to model multi-omics data in order to map and elucidate the genetic and molecular mechanisms involved in ALS, i.e., the molecular landscape building approach and the Patrimony platform. These two methods are complementary because they rely upon different omics data sets, analytic methods, and scoring systems to identify and rank therapeutic target candidates. The orthogonal combination of the two modeling approaches led to significant convergences, as well as some complementarity, both for validating existing therapeutic targets and identifying novel targets. As for validating existing targets, we found that, out of 217 different targets that have been or are being investigated for drug development, 10 have high scores in both the landscape and Patrimony models, suggesting that they are highly relevant for ALS. Moreover, through both models, we identified or corroborated novel putative drug targets for ALS. A notable example of such a target is MATR3, a protein that has strong genetic, molecular, and functional links with ALS pathology. In conclusion, by using two distinct and highly complementary disease modeling approaches, this study enhances our understanding of ALS pathogenesis and provides a framework for prioritizing new therapeutic targets. Moreover, our findings underscore the potential of leveraging multi-omics analyses to improve target discovery and accelerate the development of effective treatments for ALS, and potentially other related complex human diseases.

Keywords: MATR3; amyotrophic lateral sclerosis (ALS); molecular landscape; multi-omics modeling; patrimony; therapeutic targets.

Figure 1

Description of the ALS landscape 2.0. (a) Composition of the list of 326 ‘color-coded’ candidate genes/proteins—dark green, light green, yellow, and blue—plus an additional 14 ‘grey’ candidate genes that were prioritized for building the ALS landscape (see also Section 4). (b) Molecular landscape of ALS in a motor neuron. (c) Schematic representation of the molecular landscape of ALS, with its main functional themes. High-quality/-resolution versions of (a–c) are also provided as Figures S1, S3, and S5.

Figure 2

Scoring comparison between DTID and Patrimony. (a) Trend correlation between DTID landscape and Patrimony biorelevance scores, highlighting genes with a known genetic link with ALS. (b) Genes included in the DTID ALS landscape have significantly higher scores in Patrimony. Despite not being significant, a tendency to a higher score can be observed for the 47 potential targets identified in the DTID landscape; *** p < 0.001, **** p < 0.0001. (c) Correlation matrix between the DTID and Patrimony scores.

Figure 3

Targets worked on by the pharma industry in ALS in the landscape and Patrimony models. (a) Overlap among the targets worked on by the pharmaceutical industry, the ALS landscape genes, and the corresponding top 293 targets for the Patrimony biorelevance score. The names of the 10 targets common to all lists are displayed. (b) Top panel: genes on the Patrimony list or the DTID list have a higher proportion of targets for which drug development is ongoing. The proportion is even higher for targets on both lists. Bottom panel: no obvious difference in pharmaceutical phases can be seen between the different target lists. (c) Targets worked on by the pharma industry have significantly higher scores in Patrimony; **** p < 0.0001. (d) Active and inactive targets worked on by the pharma industry have higher scores in both the Drug target ID landscape and Patrimony platform. The highest pharmaceutical phase attained is represented by a color code.

Figure 4

SOD1, a well-known target, in the DTID landscape and Patrimony. (a) Interactions of SOD1 with other proteins in the DTID landscape; the color code corresponds to that in Figure 2a. (b) SOD1 scoring in Patrimony for generic or SOD1-related ALS. (c) Protein–protein interaction network of SOD1 constructed using the Patrimony list, with the corresponding biorelevance scores, for SOD1-related ALS.

Figure 5

MATR3, corroborated as a novel target in the DTID landscape and Patrimony. (a) Interactions of MATR3 with other proteins in the DTID landscape; the color code corresponds to that in Figure 2a. (b) MATR3 scoring in Patrimony for generic or SOD1-related ALS. (c) Protein–protein interaction network of MATR3 constructed using the Patrimony list, with the corresponding biorelevance scores, in generic ALS.

Figure 6

Flowchart that summarizes how we have combined two modeling approaches to identify and corroborate novel therapeutic targets for ALS.

Figure 7

Workflow of the DTID and Patrimony ALS methods. (a) Different steps/phases in building the molecular landscape of ALS by DTID, as further described in Section 4. (b) The Patrimony knowledge graph, built using both disease-specific and disease-agnostic data, allows for the calculation of ALS relevance scores (genetic, transcriptomic, and aggregation) for all protein-coding genes. Using the protein–protein interaction graphs, the equivalent gene environment scores are calculated as well. (c) High-level comparison between the two methods, which differ based on the data used, the analysis, and the number of genes for which a score is calculated. The stars indicate that these data types are used to select landscapes genes.