BPCPPDYSOE21 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 9: Systems Biology II

BP 9.5: Vortrag

Montag, 22. März 2021, 15:30–15:50, BPc

Towards an alphabet of random matrix models for large biological networks — •Philipp Fleig¹ and Ilya Nemenman² — ¹University of Pennsylvania, Philadelphia, USA — ²Emory University, Atlanta, USA

Biological interaction networks such as populations of neurons or amino acid sequences in proteins are critical to the functioning of any biological system. The trend of modern high-throughput experiments is to record data from a rapidly increasing number of simultaneously measured network units. Such data recorded from a biological network has characteristics of a large random matrix with hidden structures encoded in it. We present first steps towards the design of an alphabet of random matrix models to describe data of biological networks. Here, we focus on how to detect different random matrix structures in data from simple observable quantities such as pairwise correlations and the eigenvalue spectrum of the correlation matrix. Using random matrix theory we show analytically how properties of the data, such as a hidden dimensionality, are encoded in these observables. Finally, we use a neural network classifier with the observables as input to detect different types of random matrix structures in our alphabet and their hidden dimensionality in noisy data of finite size. Our approach can likely be used to model large and complex data of diverse types of biological networks.