NPIF Model-based machine learning of multi-omics data

How do cells change their gene expression to respond to a changing environment? How do we turn massive "multi-omics" data - measurements of many different kinds of molecular states in cells - to produce an accurate quantitative picture of changing gene expression patterns? This PhD project will develop artificial intelligence and machine learning methods to quantify multi-omics data, and apply them to sequencing datasets to understand how fungal cells dynamically regulate RNA expression and processing.

The project necessarily addresses the key technical problem of normalization. How do you compare counts of molecules per cell between two very different groups of cells? For example, the number of messenger RNA molecules per cell varies hugely in different growth states of the pathogenic fungus Cryptococcus neoformans. Current methods, that assume that most RNA molecules don't change in count, cannot accurately detect this variation. This project will develop rigourous methods to compare mRNA counts across growth states using external reference "spike-in" whole cells and RNAs.

How do you compare different molecular states in the same group of cells? For example, we have measurements of RNA in different conditions, and also of a sub-population of RNA that is regulated by a specific protein. The project will develop quantitative models of the RNA-protein interactions, and apply them to these measurements to understand how distinct RNAs are regulated as conditions change.