Machine learning analysis and genome-wide association studies (GWAS) identified biomarkers for the identification of several neurological diseases including myasthenia gravis (MG), as published in Frontiers in Molecular Neuroscience.

The researchers identified a number of predictors of disease as well as genetic predispositions for the different neurological diseases including MG, Alzheimer’s disease, Parkinson’s disease, and motor neuron disease. A multinomial model was able to identify samples with 88.3% accuracy and identified blood, urine, and cognitive biomarkers.

GWAS discovered several single nucleotide polymorphisms (SNPs) within genes that are shared between the neurological diseases, as well as several cancers and craniofacial disorders. The shared SNPs may indicate a similar susceptibility between these different conditions.

The biomarkers identified during the initial analysis of the study for MG include decreased albumin, cholesterol, and low-density lipoprotein (LDL) levels but increased cystatin C levels compared to controls. In the multinomial model, the model linked high alanine aminotransferase (ALT) and cystatin C and low albumin and LDL with MG.

Read more about MG testing

Interestingly, many of the neurological diseases had biomarkers linked to them that are normally associated with liver disease including ALT, alkaline phosphatase, and bilirubin. These results may possibly indicate a role for the brain-liver axis, which has previously been indicated with regard to Alzheimer’s disease, in several other neurological diseases.

In the GWAS, 70 SNPs were identified that were linked with at least 3 of the 4 neurological diseases. A total of 9 SNPs were found to be common in all 4 diseases including GAB2, LINC00290, ACO1, HLA-G, SIX1, HS6ST1, GALNT10, ONECUT1, and HLA-DRB6. Some of the SNPs identified for MG included those that are generally associated with cancer including MTUS1, LAIR2, and CBLB as well as some brain- and neuron-specific genes including CBLN4, KCNH5, ADCY8, and AUTS2.

“We propose data-driven machine learning and data exploration by GWAS as ideal first steps toward biomarker discovery for diseases of unknown etiology or currently lacking promising biomarkers,” the authors said.

“Such data-driven approaches may be extended to bench experimental work and are expected to guide dynamic detection and quantification of target druggability, in vivo demonstration of mechanisms of action, and prediction of drug resistance mechanisms,” the authors concluded.

Reference

Lam S, Arif M, Song X, Uhlén M, Mardinoglu A. Machine learning analysis reveals biomarkers for the detection of neurological diseases. Front Mol Neurosci. 2022;15:889728. doi:10.3389/fnmol.2022.889728