Claassen Lab Semester Project: Deep Learning for dynamic system motif detection in health and disease
Cells process external and internal biochemical signals by means of signaling cascades, which constitute complex, dynamical networks. In this project we use novel deep (machine) learning techniques to decompose such networks in their building blocks in order to elucidate the overall structure.
Learning the structure of dynamical, biochemical networks in health and disease from potentially incomplete and noisy time course data is a non-trivial task in computational systems biology. In this project, we assume these networks can be described as composition of smaller and repeatedly used building blocks, so-called network motifs [1], like feed-back or feed-forward loops.
Our goal is to learn models of such motifs with possibly deep neural networks from single-cell training data and to subsequently identify motif occurrences in larger data sets such as those for cancer and immunology related processes as TRAIL induced apoptosis. More specifically, we will explore the applicability of convolutional neural networks [2,3] as motif detectors. In a subsequent step, trained motif detectors will be used to identify elementary network motifs present in bigger biological networks via integration into a sparse regularized neural network.
The project will be purely computational, and would suit a student with strong interest in dynamical systems, machine learning and programming (Matlab or Python), preferably with a background in computational biology, computer science, mathematics or statistics.
For further information, please contact Professor Manfred Claassen (), Eirini Arvaniti () or Stefan Ganscha ()
[1] Alon, U. (2007). Network motifs: theory and experimental approaches. Nature Reviews. Genetics, 8(6), 450–61.
[2] LeCun, Y., Bengio, Y. & Hinton, G. (2015). Deep learning. Nature 521, 436–444.
[3] Alipanahi, B., Delong, A., Weirauch, M. T., & Frey, B. J. (2015). Predicting the sequence specificities of DNA-and RNA-binding proteins by deep learning. Nature biotechnology.