INVITED SPEAKERS

 

Assoc. Prof. Roberto C. Sotero
University of Calgary, Canada

 

Roberto C. Sotero earned his BSc in Nuclear Physics in 2003 and his PhD in Physics in 2009, both from Havana, Cuba. He then pursued a postdoctoral fellowship at the Montreal Neurological Institute from 2010 to 2014. In 2014, he joined the Department of Radiology at the University of Calgary, where he currently holds the position of Associate Professor. Roberto's expertise lies in developing computational models of brain activity, bridging gaps in imaging across various scales and between brain structure and function. His work focuses on the essential biophysical properties that influence empirical dynamics. Recently, he began integrating machine learning techniques with biophysical modelling. Through these hybrid models, he aims to address critical health challenges, particularly the early detection of autism spectrum disorder.

 

Speech Title: "Advancing Brain Connectivity Analysis: Physics-Informed Neural Networks for Neuroimaging Data"

 

Abstract: Understanding the intricate dynamics of brain connectivity is essential for unraveling the neural mechanisms underlying cognitive functions, states of consciousness, and neurological disorders. In this work, we present two innovative approaches leveraging Physics-Informed Neural Networks (PINNs) to address key challenges in brain connectivity analysis. First, we introduce a novel framework for estimating directed functional connectivity (dFC) from resting-state functional magnetic resonance imaging (rsfMRI) data. By integrating a brain dynamics model (BDM) with PINNs, we simultaneously estimate biophysical parameters and dFC across brain regions. Our method overcomes traditional limitations in accuracy and scalability, revealing significant sex-specific differences in connectivity patterns in both neurotypical individuals and those with autism spectrum disorder (ASD). These findings underscore the importance of directionality in understanding brain networks and their role in neuropsychiatric conditions. Second, we extend the PINN framework to estimate the excitatory/inhibitory (E/I) balance from electrocorticography (ECoG) data. By embedding a neural mass model (NMM) into the PINN architecture, we estimate local and long-range connectivity parameters under both resting and anesthesia conditions. Our results demonstrate a significant reduction in long-range connections and excitatory short-range connections, alongside an increase in inhibitory short-range connections under anesthesia. These findings provide insights into how anesthetic agents modulate neural dynamics to induce unconsciousness, corroborating existing theories on the neural mechanisms of anesthesia. Together, these studies highlight the power of PINNs in advancing our understanding of brain connectivity. By integrating physical laws with data-driven learning, our approaches offer robust, interpretable, and scalable solutions for estimating both directed functional connectivity and the E/I balance. These advancements not only deepen our understanding of neural dynamics but also hold promise for improving diagnostic tools and therapeutic interventions in neuroscience.

 

 

Assoc. Prof. Sung-Joon Park

The University of Tokyo, Japan

 

Sung-Joon Park holds a PhD in Engineering from Tokyo Institute of Technology, which he received in 2005. After his doctoral studies, he pursued research positions at Kobe University and Kyoto University. In 2010, he joined the Human Genome Center at the Institute of Medical Science, the University of Tokyo, where he was appointed associate professor in 2020 (https://www.hgc.jp/~park/). Sung-Joon Park's primary research focus is in computational biology, where he investigates the genomic and epigenomic impacts across various biological and medical contexts. His work is particularly notable for the development of innovative computational approaches and databases that aim to advance our understanding of genetics and biomedical sciences. He is a member of Information Processing Society of Japan (IPSJ), the Japan Society for Artificial Intelligence (JSAI), and Japanese Society for Bioinformatics (JSBi).

 

Speech Title: "A Graph-Embedding Approach to Dissecting Proximal and Distal Gene Regulators"

 

Abstract: The spatial organization of the genome plays a critical role in mediating the functional effects of distal chromosomal interactions. In particular, enhancer-promoter interactions have been intensively studied using advanced computational algorithms. However, our understanding of how enhancer signals are transmitted to their target promoters through complex regulatory networks remains limited. In this study, we developed a novel computational framework that combines a regression model, which predicts gene expression by identifying key promoter-distal and -proximal regulators, with a graph-embedding algorithm designed to detect cell-type-specific and conserved regulatory interactions within complex gene regulatory networks. We applied this method to human naïve and germinal center B cells and, as a result, identified sets of promoter-distal transcription factors and architectural cofactor proteins, which are co-regulated to maintain cellular stability and prevent malignancy. These findings emphasize the importance of understanding both cis- and trans-regulatory interactions in the transcriptional machinery. Our approach provides a valuable alternative for studying enhancer biology and its mediation through protein-protein interactions within the context of 3D genome organization.