is a Joint CS and ECE seminar.
In order to feed the ever-growing population, it is important to produce food as efficiently as possible. However, farmers have difficulty predicting the best time to harvest their crops when weather patterns fluctuate, as they have in recent years. In order to improve our agricultural capability, we need to better understand how plants sense their environment and how they integrate this information into their growth and development programmes. I will discuss three projects from my research group that illustrate how we map the interactions between genes, environments and phenotypes, using various statistical and computational strategies.
First, I will discuss a project to understand how environmental factors influence phenotype. We ran a citizen science campaign in British primary schools to collect detailed time series data of spring onion growth under naturalistic temperature and light fluctuations, and then we used functional regression strategies to model how plants respond to environmental signals in a time-dependent way.
The second project helped us understand the relationship between gene expression and phenotype. While there is a lot of RNA-seq data that is publicly available on international databases, it is difficult to utilise this data for meta-analysis because phenotypes are not well-annotated and there could be substantial batch effects across research labs. We have developed a tool called TissueTimer to help determine whether observed variations in gene expression might be a consequence of different tissue ratios or sample ages, using Bayesian statistics and kernel methods.
The third project focused on how environmental fluctuations are integrated by gene networks. In particular, we were interested in unravelling the gene expression dynamics that occur at dawn, a time period in which genes change their expression levels extremely rapidly. In order to avoid missing interesting dynamics, we needed to select time points for our study that would be as informative as possible, so we developed NITPicker, a time point selection tool that utilised functional data analysis and a Viterbi-like dynamic programming algorithm to select informative time points. The resulting data enabled us to infer gene regulatory networks that could explain the gene expression patterns observed in the early morning.
Taken together, these projects illustrate how environmental signals are integrated by gene regulatory networks and how this influences plant growth and development.
Dr. Daphne Ezer is currently a Research Fellow at the Alan Turing Institute for Data Science in the British Library in London and the University of Warwick Department of Statistics. She was previously a research associate in the Sainsbury Laboratory in the University of Cambridge working with Dr. Philip Wigge, and a PhD student in the University of Cambridge System Biology centre supervised by Dr. Boris Adryan. She is interested in using data science to unravel how genes are regulated in plants, in order to inform agricultural practice.