intro

Our lab develops machine learning and probabilistic methods that connect genetic variation to biological function, at two scales.

At population scale, we model human genetic variation — ancestry, recombination history and admixture — to understand how inherited variation shapes phenotype across diverse populations, including in biobank-sized cohorts.

At cellular and neural scale, we work with multimodal omic data (e.g., scRNA-seq) and functional recordings (e.g., Ca imaging) to link molecular programs to brain function and disease.

What ties the two together is methodological: deep generative modeling, probabilistic inference and representation learning applied to large, noisy, multimodal biological data. A recurring motivation is to make in silico models good enough to answer questions that would otherwise require animal experimentation.

research areas

Our work develops machine learning and probabilistic methods to connect genetic and molecular variation with cellular state, brain function and behavior. Current projects span population and statistical genetics, multi-omics integration, neural data foundation models, and clinically motivated genomics and neuroscience applications.

population and statistical genetics

• Ancestry inference and admixture mapping in diverse and admixed cohorts
• Learning on ancestral recombination graphs and genome-wide genealogical structure
• Biobank-scale association and phenotype prediction, with attention to multi-ancestry generalization
• Open-source tooling for genetic variation and population structure

genotype–phenotype modeling and genomic AI

• Genotype-to-phenotype prediction, from neurodegeneration to complex traits at biobank scale
• Sequence-based representation learning (genomic language models) and multimodal genomic prediction
• Projects linking inherited variation, molecular readouts, and downstream phenotypes, including telomere-related traits

single-cell and multi-omics genomics

• Computational pipelines for single-cell and multi-omics data analysis and QC
• Multimodal integration and representation learning across omics layers (for example transcriptomics, proteomics, and epigenomics)
• Latent factor and trajectory methods to map cellular programs and disease-associated perturbations, including exposure and mental health contexts

computational neuroscience and neural foundation models

• Deep generative and probabilistic models for neuronal functional recordings, with emphasis on calcium imaging
• Single-neuron and population-level dynamics, simulation-based validation, and scalable inference
• Foundation-model approaches for neural data and neurotechnology, including photostimulation-oriented modeling

integrative neurogenomics and translational brain research

• Multi-modal integration of molecular and functional data to study brain-related phenotypes
• Collaborative projects combining omics and neuroscience data in disease and intervention contexts, including multi-partner networks

environment

Our lab is part of various world-class research institutions and groups in Barcelona, including:

• Institute of Neurosciences at UB (UBNeuro) [URL]
• TECNIO Group of Data Science and Artificial Intelligence at UB [URL]
• Group of Pathophysiology and Treatment of Neurodegenerative Disorders at IDIBAPS [URL]

We also collaborate closely with the Ioannidis lab at Stanford University and UC Santa Cruz, with whom we co-supervise students and develop much of our population-genetics work.

Our lab’s research pushes the boundaries of what can be achieved with in silico models in the life sciences, from population genomics to neuroscience, and we are excited to be at the forefront of discoveries with the potential to transform both. We welcome collaborations and discussions with others interested in our work.