We investigate data science problems. Our research stream integrates theory and experiment and merges data analysis with data acquisition. We focus on open questions in biophysics, biochemistry, and biomedicine. Overall, our work is cross-disciplinary and relies on: high efficiency measurements, high fidelity modeling, and high performance computing.
The areas of our expertise include numerical and Bayesian computing, mathematical biology as well as computational aspects of biophysics/biochemistry and biomedicine.
See our publications, recent and past, on Google Scholar and ORCID.
Get in touch with our group firstname.lastname@example.org, 323 Ayres Hall, UTK.
Follow us on Twitter @SgouralResearch.
Join our group
We are currently recruiting graduate and undergraduate students for research projects in Data Science/Math Bio/Applied Math. For information contact Dr Sgouralis at email@example.com.
Data science in the natural and life sciences
We are interested in formulating and analyzing complex dynamical systems of living organisms from the whole body to the single molecule level. We adapt tools and notions from applied mathematics, computer science, and statistics to solve problems in the natural and life sciences. We develop computational methods to characterize interactions of biological mechanisms using noisy and often incomplete data and we focus on raw datasets acquired in state-of-the-art experiments.
Single molecule experiments
Modern experiments probe space and time with enough sensitivity to reveal the dynamics of individual bio-molecules. For example, imaging techniques such as super-resolution microscopy enable the localization of individual bio-molecules, fluorescence FRET or FCS measurements reveal the dynamics of protein conformational changes or fast molecular motion, and force spectroscopy utilizing optical tweezers can monitor forces as weak as those exerted by single proteins or small nucleic acids.
As these methods can assess functioning bio-molecules, often in their natural environment, they provide unique means of probing molecular interactions and elucidate fundamental biochemical processes. However, despite the abundance of experimental data, interpreting the acquired data in physical terms and drawing meaningful insights is severely limited by fundamental challenges pertaining to the nature of the systems at hand: in vivo or in vitro biological processes are obscured by stochastic events, detector devises introduce excessive noise, and experimental artifacts such as photo-physics or protocol specific characteristics often mask the bio-molecular dynamics of interest.
To address these challenges, we develop specialized analysis methods to interpret raw measurements and derive statistical models of complex biological systems directly from experimental data. Our models often combine Bayesian nonparametric Statistics and extensive Markov chain Monte Carlo computing.
Image analysis and tracking
In the sub-cellular level, routinely performed fluorescence microscopy reveals individual particles, such as organelles or even individual bio-molecules. Since fluorescent imaging is commonly performed in a sequential manner and several successive images of the same structures are obtained in sort time periods, the spatio-temporal behavior of the fluorescent particles is readily captured and subsequently may be revealed with proper analysis. However, despite extensive utilization of fluorescent imaging and the availability of vast datasets, characterizing the spatio-temporal dynamics captured is severely limited. For instance, limiting factors include: highly noisy measurements contaminated with optical aberrations caused by either the environments in which the particles move of the detector equipment, large number of moving particles, high volume of imaging datasets, and most importantly unknown motion models.
To address these challenges, we develop computational algorithms to analyze raw measurements and characterize complex motion patterns from imaging data. As a fundamental limitation lies on the absence of motion model, we adapt novel concepts from topological data analysis with velocimetry for motion reconstruction.
We also specialize on single molecule imaging experiments and multi-target tracking applications capable of analyzing super-resolution microscopy datasets. Our approaches aim to extract spatio-temporal characteristics of single molecule motion from wide-field or confocal microscopy while simultaneously characterizing motion patterns and individual trajectories.
Biological systems across scales
Mechanistic modeling of large scale biological systems, such as cellular, organ, or even whole body systems, can combine information from multiple sources and summarize an extensive amount of relevant knowledge. Often, such comprehensive models are necessary to investigate complex biomedical conditions in either health or disease or even to characterize the patho-physiology of clinical interventions. As most of these conditions cannot be investigated directly by experimental or clinical means, properly constructed mathematical models combining current understanding and fusing physical, experimental, and clinical findings can elucidate complex interactions that are inaccessible otherwise.
Our research is focused on the mathematical formulation and analysis of such systems. We are particularly interested in investigating the properties and behavior of the cardiovascular and urinary systems. Our focus is on the representation and quantification of hemodynamic interactions between the kidney’s autoregulatory mechanisms, transport mechanisms, and tissue oxygenation. Our interests include the elucidation of multi-scale phenomena pertaining to the auto-regulatory response of the renal vasculature, the supply of the renal medulla, and the changes induced by open heart surgery.