Identification of Parkinson's disease PACE subtypes and repurposing treatments through integrative analyses of multimodal data

Abstract

Parkinson's disease (PD) is a serious neurodegenerative disorder marked by significant clinical and progression heterogeneity. This study aimed at addressing heterogeneity of PD through integrative analysis of various data modalities. We analyzed clinical progression data (≥5 years) of individuals with de novo PD using machine learning and deep learning, to characterize individuals' phenotypic progression trajectories for PD subtyping. We discovered three pace subtypes of PD exhibiting distinct progression patterns: the Inching Pace subtype (PD-I) with mild baseline severity and mild progression speed; the Moderate Pace subtype (PD-M) with mild baseline severity but advancing at a moderate progression rate; and the Rapid Pace subtype (PD-R) with the most rapid symptom progression rate. We found cerebrospinal fluid P-tau/α-synuclein ratio and atrophy in certain brain regions as potential markers of these subtypes. Analyses of genetic and transcriptomic profiles with network-based approaches identified molecular modules associated with each subtype. For instance, the PD-R-specific module suggested STAT3, FYN, BECN1, APOA1, NEDD4, and GATA2 as potential driver genes of PD-R. It also suggested neuroinflammation, oxidative stress, metabolism, PI3K/AKT, and angiogenesis pathways as potential drivers for rapid PD progression (i.e., PD-R). Moreover, we identified repurposable drug candidates by targeting these subtype-specific molecular modules using network-based approach and cell line drug-gene signature data. We further estimated their treatment effects using two large-scale real-world patient databases; the real-world evidence we gained highlighted the potential of metformin in ameliorating PD progression. In conclusion, this work helps better understand clinical and pathophysiological complexity of PD progression and accelerate precision medicine.

Fig. 1. A diagram illustrating the present analysis.

a Collecting longitudinal clinical data from the Parkinson’s Progression Markers Initiative (PPMI) and Parkinson’s Disease Biomarkers Program (PDBP) cohorts and conducting necessary data cleaning and preprocessing. b Development of a deep phenotypic progression embedding (DPPE) model to learn a progression embedding vector for each individual, which encodes his/her PD symptom progression trajectory. c Cluster analysis with the learned embedding vectors to identify PD subtypes, each of which reveal a unique PD progression pattern. d Identifying CSF biomarkers and imaging markers the discovered PD subtypes. e Construction of PD subtype-specific molecular modules based on genetic and transcriptomic data, along with human protein-protein interactome (PPI) network analyses, using network medicine approaches. f In silico drug repurposing based on subtype-specific molecular profiles and validation of drug candidates’ treatment efficiency based on analysis of large-scale real-world patient databases, i.e., the INSIGHT and OneFlorida + . g Architecture of the DPPE model. Specifically, DPPE engaged two Long-Short Term Memory (LSTM) units—one as encoder receiving an individual’s longitudinal clinical records and compacting them into a low-dimensional embedding space; while another taking the individual’s embedding vector to reconstruct the original clinical records. DPPE was trained by minimizing the reconstruction difference.

Fig. 2. Progression patterns of the three PD subtypes within the PPMI cohort.

a Averaged progression trajectories in clinical manifestations by subtypes, with shading indicating standard error of the mean (SEM). b Sankey diagrams showing evolution patterns of motor phenotypes (tremor dominant, indeterminate, and PIGD) by subtypes. c Sankey diagrams showing evolution patterns of cognition phenotypes (normal cognition, MCI, and dementia) by subtypes. d Sankey diagrams showing evolution patterns of mood phenotypes (normal, mild depression, moderate depression, and severe depression) by subtypes. e Sankey diagrams showing evolution patterns of sleep phenotypes (REM sleep behavior disorder [RBD] negative and positive) by subtypes.

Fig. 3. CSF biomarkers and neuroimaging markers of the identified subtypes.

a CSF biomarkers by PD subtypes. On each box plot, the central mark indicates the median value and the bottom and top edges of the box indicate the interquartile range (IQR) with whiskers covering the most extreme values within 1.5 × IQR. b Regions showing significant signals in 1-year brain atrophy between a pair of subtypes. 1-year brain atrophy was measured by cortical thickness and white matter volume from 34 region of interests (ROIs), defined by the Desikan-Killiany atlas (averaged over the left and right hemispheres). Color density denotes significance in terms of -log₁₀(P).

Fig. 4. PD-R subtype-specific molecular modules revealing potential biological mechanisms of rapid PD progression.

a Genetic molecular module of PD-R. b Pathways enriched based on genetic molecular module of PD-R. c A sub-network of transcriptomic molecular module of PD-R. The entire transcriptomic molecular module of PD-R can be in the Supplementary Fig. 9. d Pathways enriched based on transcriptomic molecular module of PD-R.

Fig. 5. Identified repurposable drug candidates for preventing PD progression by targeting subtype-specific molecular changes.

a Gene set enrichment analysis (GSEA) based on subtype-specific gene modules with bulk RNA-seq data of individuals and transcriptomics-based drug-gene signature data in human cell lines identified repurposable drug candidates for different PD pace subtypes. Treatment effect estimation using the INSIGHT data within the broad PD population (b) and probable PD-R population (c). Treatment effect estimation using the OneFlorida+ data within the broad PD population (d) and probable PD-R population (e). ^aThe drug doesn’t have sufficient patient data (<100) for analysis. ^bThe drug does not have sufficient balanced emulated trials (<10). N_T indicates the number of eligible PD patients who received the tested drug after PD initiation.

Fig. 6. Comparisons of the identified pace subtypes with conventional motor subtypes and prior data-driven subtypes.

Notably, our subtyping algorithm was completely data-driven and hypothesis-free. In addition, since our method modeled individuals’ phenotypic progression profile for PD subtyping, the identified subtypes demonstrated unique progression patterns and, importantly, were stable over time.