Explore projects

Graph representation learning modelling pipeline exploiting molecular interaction networks of transcriptomics (protein-protein interactions) and metabolomics (metabolite-metabolite interactions) to learn PD-specific fingerprints from the spatial distribution of molecular relationships in an end-to-end fashion. The scripts apply the graph representation learning modelling pipeline on networks of molecular interactions, where transcriptomics and metabolomics data from the PPMI and the LuxPARK cohort, respectively, are projected.

This repository contains the code for ML analyses performed in Chapter 4 of my PhD thesis "Interpretable Machine Learning on omics data for biomarker discovery in Parkinson's disease". The project consists on performing Parkinson's disease (PD) case-control classification from blood plasma metabolomics measurements at the baseline clinical visit from the LuxPARK cohort, and from whole blood transcriptomics data at baseline as well as dynamic features engineered from a short temporal series of 4 timepoints from the PPMI cohort. The study involves evaluation of different feature selection strategies, The goal was to build and test a collection of ML models and, most interestingly, identify molecular and higher-level functional representations associated with PD diagnosis.

This repository contains the code for ML analyses performed in Chapter 5 of my PhD thesis "Interpretable machine learning on omics data for the study of UPDRS III prognosis". The project consists on predicting the Unified Parkinson’s Disease Rating Scale Part III (UPDRS III) motor scores (mild/severe when classification) from whole blood transcriptomics and blood plasma metabolomics using measurements from the baseline clinical visit, and temporal or dynamic features engineered from a short temporal series of 4 and 3 timepoints, respectively, from the PPMI cohort and the LuxPARK cohort, aiming at identifying molecular and higher-level functional fingerprints linked specifically to the motor symptoms in PD.

This repository contains the code for statistical analyses performed in Chapter 3 of my thesis "Cross-sectional and longitudinal profiling of PD transcriptomics and metabolomics". The project consists on whole blood transcriptomics and blood plasma metabolomics cross-sectional and longitudinal profiling of Parkinson's disease patients and controls from the PPMI cohort and the LuxPARK cohort respectively, to identify differential molecular and higher-level functional features in PD.

A package and environment with the NestedCV() class - a scikit-learn compatible class for nested cross-validation.

Code related to Talavera Andujar et al DOI: 10.26434/chemrxiv-2023-6j2gm

ShinyTPs: Curate transformation products from text mining results from HSDB & PubChem