A Knowledge-Based Machine Learning Approach to Gene Prioritisation in Amyotrophic Lateral Sclerosis

Abstract

Amyotrophic lateral sclerosis is a neurodegenerative disease of the upper and lower motor neurons resulting in death from neuromuscular respiratory failure, typically within two to five years of first symptoms. Several rare disruptive gene variants have been associated with ALS and are responsible for about 15% of all cases. Although our knowledge of the genetic landscape of this disease is improving, it remains limited. Machine learning models trained on the available protein–protein interaction and phenotype-genotype association data can use our current knowledge of the disease genetics for the prediction of novel candidate genes. Here, we describe a knowledge-based machine learning method for this purpose. We trained our model on protein–protein interaction data from IntAct, gene function annotation from Gene Ontology, and known disease-gene associations from DisGeNet. Using several sets of known ALS genes from public databases and a manual review as input, we generated a list of new candidate genes for each input set. We investigated the relevance of the predicted genes in ALS by using the available summary statistics from the largest ALS genome-wide association study and by performing functional and phenotype enrichment analysis. The predicted sets were enriched for genes associated with other neurodegenerative diseases known to overlap with ALS genetically and phenotypically, as well as for biological processes associated with the disease. Moreover, using ALS genes from ClinVar and our manual review as input, the predicted sets were enriched for ALS-associated genes (ClinVar p = 0.038 and manual review p = 0.060) when used for gene prioritisation in a genome-wide association study.

Genome-wide Meta-analysis Finds the ACSL5-ZDHHC6 Locus Is Associated with ALS and Links Weight Loss to the Disease Genetics

“…Such results, considering the role of G2E3 in the regulation of DDR, suggest that it could play an important role in the development of ALS. SCFD1 is also involved in vesicle transport (Hou et al, 2017); we previously reported the association between ALS and SCFD1 SNPs in a European GWAS using linear mixed model analysis (van Rheenen et al, 2016) and in our recently developed machine learning method for gene discovery in ALS (Bean et al, 2020). However, our attempts in both the same GWAS (van Rheenen et al, 2016) and successive GWASs (Nicolas et al, 2018;Benyamin et al, 2017) failed to replicate it.…”

“…ATXN3 expression controls and is essential for the recruitment of mutated SOD1 into toxic aggresomes (Wang et al, 2012), one of the most common causes of ALS. Also, ATXN3 was predicted by our machine learning method (Bean et al, 2020). TRIP11 encodes for a protein associated with the Golgi apparatus and is involved in vesicle transport (Follit et al, 2008).…”

What Can Machine Learning Approaches in Genomics Tell Us about the Molecular Basis of Amyotrophic Lateral Sclerosis?

“…A popular strategy that has been used to select the initial data and then further reduce the feature and instance space is ALS-associated knowledge (see Table 1 ). The most relevant study that falls into this category is Bean et al which modelled only previously known ALS-linked gene lists, mining information from the literature and from disease databases, as well as including a manually curated set of ALS-associated genes [ 105 ]. In order to reduce the feature space they performed an enrichment test on all features and, for the predictive model, kept only those features that are significantly enriched in the mechanism(s) of the disease [ 116 ].…”

“…Another example is Yin et al which before applying their Promoter-CNN model as a feature selection method, they first limited their feature space by studying only non-additive phenomena of multiple promoters located on four specific chromosomes (7, 9, 17 and 22), those chromosomes have being selected based on the amounts of missing heritability that have been previously identified in ALS [ 36 , 90 ]. Bean et al, Kim et al, Yousefian et al and Sha et al, implement multi- step algorithms in which one of the initial steps reduces the feature space by keeping only the highest performing genes/SNPs, by assessing the enrichment of genes in ALS [ 105 ] or by applying a specific threshold of single-marker association analysis to ALS [ 91 , 104 , 108 ].…”

“…Another study that combined genomic data with other types of data sources to increase the power of the machine learning gene prioritization method was that of Bean et al, integrating functional annotations, known ALS-gene associations, and protein-protein interactions [ 105 ]. Protein-protein interaction networks have been useful to decipher new disease mechanisms, as proteins that are encoded by disease-related genes are likely to interact with proteins that are implicated in similar pathologies [ 135 ].…”