The intrinsic geometry of a developing embryo
A developing embryo such as the Drosophila melanogaster is a complex system involving dynamics at multiple scales, from single molecules, to cells, to tissues, to organs. The recent years have witnessed a tremendous development in measurement techniques for developmental biology, from in toto imaging to single cell RNA sequencing. Each of these measurement techniques brings out unique features of a developing embryo (cell differentiation, morphogenesis, ..) leading to the generation of large amounts of heterogeneous data. There is therefore a need to integrate them into a common representation to extract and understand correlations between multiple scales. On the other hand, machine learning methods such as manifold learning or deep learning offer unprecedented ways of fusing heterogeneous data by learning mappings between high dimensional spaces from empirical data, therefore opening new ways for quantitative integrative developmental biology.
We want to develop a general framework for data integration in developmental biology using the fly Drosophila melanogaster as a model system. We will take advantage of the intrinsic geometry of the data to integrate multiple views of a developing embryos from various experimental systems. The various specificities of a developing embryo such as its time dependency and its spatial organization provide constraints that can be taken advantage of when designing algorithms for data integration. Various structures can be extracted from the data such as low dimensional manifolds and be used to find relations between heterogeneous measurements.
We propose to first develop a general machine learning methodological framework for data integration in a developing embryo seen as a complex system. We will then design specific algorithms and test them on open data sets of live microscopy and single cell RNA sequencing data. This should lead to the development of a library that could be used by many researchers around the world. This work could pave the way for collaboration with experimental laboratories at the Turing Centre for Living Systems.
PhD student’s expected profile
The student should have a background in machine learning and a good knowledge of biology as well as a will to open new avenues at the intersection of biology and computer science.