Multi-modality machine learning predicting Parkinson’s disease

Abstract

Personalized medicine promises individualized disease prediction and treatment. The convergence of machine learning (ML) and available multimodal data is key moving forward. We build upon previous work to deliver multimodal predictions of Parkinson’s disease (PD) risk and systematically develop a model using GenoML, an automated ML package, to make improved multi-omic predictions of PD, validated in an external cohort. We investigated top features, constructed hypothesis-free disease-relevant networks, and investigated drug–gene interactions. We performed automated ML on multimodal data from the Parkinson’s progression marker initiative (PPMI). After selecting the best performing algorithm, all PPMI data was used to tune the selected model. The model was validated in the Parkinson’s Disease Biomarker Program (PDBP) dataset. Our initial model showed an area under the curve (AUC) of 89.72% for the diagnosis of PD. The tuned model was then tested for validation on external data (PDBP, AUC 85.03%). Optimizing thresholds for classification increased the diagnosis prediction accuracy and other metrics. Finally, networks were built to identify gene communities specific to PD. Combining data modalities outperforms the single biomarker paradigm. UPSIT and PRS contributed most to the predictive power of the model, but the accuracy of these are supplemented by many smaller effect transcripts and risk SNPs. Our model is best suited to identifying large groups of individuals to monitor within a health registry or biobank to prioritize for further testing. This approach allows complex predictive models to be reproducible and accessible to the community, with the package, code, and results publicly available.

Establishing gene regulatory networks from Parkinson’s disease risk loci

“…7 ). The co-location of genes and eQTLs, identified as being important for Parkinson’s disease diagnosis, 38 within the nine clusters supports the potential importance of the gene–gene interactions and enriched pathways in Parkinson’s disease.…”

“…Makarious et al 38 recently used a multimodal machine learning approach, incorporating multi-omics datasets, to inform and improve predictions of Parkinson’s disease. Beyond the 90 GWAS SNP signals (which collectively were the top genetic feature), they also identified rs10835060 and rs4238361 as two SNPs that impact on Parkinson’s disease biology.…”

“…Makarious et al 38 recently utilized a multimodality approach to identify genetic and transcriptomic features that contribute to risk predictions of Parkinson’s disease. They highlighted two SNPs (rs10835060 and rs4238361) and 29 genes.…”

Changes in the cerebrospinal fluid proteome precede and stratify the course of Parkinson’s Disease

“…Recent pioneering studies have reported that combining multiple pieces of information, including disease-associated protein levels, the presence of α-synuclein seeds, genomics, transcriptomics, and clinical information, can improve the accuracy of diagnosis and/or stratification. 75–79 However, previous studies have rarely succeeded in improving all aspects of diagnosis, motor symptom stratification, and cognitive symptom stratification. 80 It has also been reported that the polygenic susceptibility that drives the onset of PD and the progression of symptoms may not be the same.…”

Polygenic Resilience Modulates the Penetrance of Parkinson Disease Genetic Risk Factors

“…Source code and documentation are available at https://genoml.com/ and on GitHub (https://github.com/GenoML/genoml2). The process we used for selecting the best p value threshold mirrored that in Makarious et al 22 We ran the discrete supervised learning workflow for munging, followed by training on a series of p value thresholds taken from Chang et al 18 PD GWAS summary statistics, which included each incremental order of magnitude ranging from 0.01 to 1 × 10 −8 . Each model included SNPs as well as sex, age, and 20 principal components (PCs) to account for population stratification.…”