I work at the intersection between molecular metabolism and computational biology. My skill-sets are within the areas of molecular biology and computer science so I integrate them to understand the deep underlying order within complex biological systems. On the one hand, I apply machine learning algorithms for pattern discovery in biological data, followed by experimental validation. On the other hand, I'm interested in finding ways to emulate strategies from evolutionary molecular adaptations to improve existing machine learning systems.

Metabolism and non-equilibrium compartmentalization in living and machine learning systems

(In collaboration with Dr. Remy Tuyeras from the MIT Mathematics and CSAIL Department)

• Transcriptional condensates as topological integrative models for transcriptional control

• Cellular intelligence: dynamic specialization through non-equilibrium multi-scale compartmentalization

• Dissipative compartmentalization drives robust specialization in biological and machine learning systems

Topological integrative biology for systems genomics

• Geometric machine learning in biological networks: Geometric machine learning is on the rise and its application to relational graphs is outstanding. Among the latter, biological networks represent one of the fields that could benefit the most from this emerging technique. Therefore, we use and develop models that combine network theory and geometric machine learning to integrate multi-omic information. Briefly, in order to achieve combinatorial generalization of biological systems, we aim to systematically learn structured representations such as networks using relational collective behavior. These methods support relational reasoning and generalization, which will help us understanding structured relations such as those observed in complex biological networks. The systematic combination of several principles of machine learning and network science will uncover relationships useful for drug-development and biomarker discovery.

• Multi-omic integrative models to study metabolic adaptations: Metabolic diseases are rarely caused by isolated events but heavily depend on concerted anomalies in high-dimensional processes. For this reason, we use and develop hybrid machine learning models suitable to study complex metabolic disease. This approach will help us to infer molecular components across cellular metabolic adaptations to energy-rich and energy-deficient environments. This systemic genomic integration allows us to infer regulatory networks governing physiological adaptations during obesity or physical exercise. 

• Network-based framework to study cell-cell and organ communication: We use tissue-specific and tissue-naïve networks containing functional biological information to study cell-cell and organ communication during metabolic disease. We then apply multi-omic data from murine models of obesity to predict organ-organ relationships during the onset of insulin resistance. We further categorize these dynamic networks by integrating genes and regulatory elements associated with type-2 diabetes and other metabolic traits. Finally, we derive hierarchical maps containing network-drivers by incorporating experimentally validated associations from (25 million articles) Medline literature. This network-based approach allows us to infer molecular variations across several organs and repurpose drug-target identification.