Grant: Discovery Grants Program - Individual
Université de Montréal
$81,000.00
Geometry of biophysical learning
FY 2023-24
Summary
Cellular processes are extremely complex and intricate, with a forest of molecular interactions emerging into high level biological functions, so that it appears impossible at first sight to extract any meaningful unifying principle in the form of a simplified mathematical description. This is in contrast with classical views from before molecular biology, such as the famous Waddington landscape in developmental biology. Indeed, despite this daunting biological complexity, many theoretical and experimental works have revealed in multiple case studies that many degrees of freedom can in fact be safely eliminated without impacting our understanding of biological function. This is done with the help of simple, geometric descriptions of biological dynamics in low dimensional spaces. As an example, we recently showed how a high dimension cytokine T cell response can be projected onto a simple bidimensional latent space. There, the biological dynamics can be studied, leading to a universal description of immune encoding (Achar et al, Science, 2022). However, it is still neither clear if such reduction can be systematically done, nor why it is possible in the first place. My research program aims at building theoretical foundations for such reduction, using multiple tools from maths, physics and computer science (including machine learning). I will further explore such how low dimensional dynamics might spontaneously emerge as a consequence of learning and evolution. First, I will study how the low dimensional dynamics we see in biological decision is also observed in artificial neural networks, and spontaneously emerge during learning. Second, I will further integrate evolutionary thinking into those descriptions, by proposing a new evolutionary algorithm to study how geometry of cellular decision can evolve, and how we can reconstruct 'Waddington' landscapes from data. Lastly, I will extend this approach to oscillatory patterns in embryonic development, to establish more examples of low dimensional description of biological dynamics.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-03843
Name nom
François, Paul
Department département
Biochimie et médecine moléculaire - Biochimie et médecine moléculaire
Organization id
63
Institution établissement
Université de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
81000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1505
Committee name en
Physics
Committee name fr
Physique
Area of application code
1202
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Physical sciences
Area of application fr
Sciences physiques
Research subject code
9999
Research subject group en
Not available
Research subject group fr
Not available
Research subject en
Non disponible
Research subject fr
Non disponible
Application title
Geometry of biophysical learning
Keyword
cellular decision; developmental biology; dynamical systems; gene networks; genetic algorithms; hopfield networks; information theory; machine learning; mathematical modelling; systems biology
Application summary
Cellular processes are extremely complex and intricate, with a forest of molecular interactions emerging into high level biological functions, so that it appears impossible at first sight to extract any meaningful unifying principle in the form of a simplified mathematical description. This is in contrast with classical views from before molecular biology, such as the famous Waddington landscape in developmental biology. Indeed, despite this daunting biological complexity, many theoretical and experimental works have revealed in multiple case studies that many degrees of freedom can in fact be safely eliminated without impacting our understanding of biological function. This is done with the help of simple, geometric descriptions of biological dynamics in low dimensional spaces. As an example, we recently showed how a high dimension cytokine T cell response can be projected onto a simple bidimensional latent space. There, the biological dynamics can be studied, leading to a universal description of immune encoding (Achar et al, Science, 2022). However, it is still neither clear if such reduction can be systematically done, nor why it is possible in the first place. My research program aims at building theoretical foundations for such reduction, using multiple tools from maths, physics and computer science (including machine learning). I will further explore such how low dimensional dynamics might spontaneously emerge as a consequence of learning and evolution. First, I will study how the low dimensional dynamics we see in biological decision is also observed in artificial neural networks, and spontaneously emerge during learning. Second, I will further integrate evolutionary thinking into those descriptions, by proposing a new evolutionary algorithm to study how geometry of cellular decision can evolve, and how we can reconstruct 'Waddington' landscapes from data. Lastly, I will extend this approach to oscillatory patterns in embryonic development, to establish more examples of low dimensional description of biological dynamics.