Optimisation of a Statistical Framework for Locating RNA modifications from Nanopore Sequencing Data

Optimisation of a Statistical Framework for Locating RNA modifications from Nanopore Sequencing Data

Statistics & Mathematics

Jessie (Jie) Zhao

Professor Anastasia Papavasiliou and Professor Jon Forster

How can we accurately detect and locate a modification on an RNA molecule using nanopore sequencing data? This is an open problem that requires ideas from both statistics and biology. Supervised by Professors Anastasia Papavasiliou and Jon Forster, I began by simulating the signal data from Oxford Nanopore devices, where a sliding window reads five bases on the mRNA chain at a time. Using R, we constructed a statistical framework which involves working with different distributions such as poisson, inverse gamma and the normal distribution to generate signal values. Then we analysed our signal values, performing hypothesis tests and calculating maximum likelihood scores in order to determine where modifications are likely to be on the mRNA chain. This year, after gaining valuable biological insight at the Laboratory of Immune Diversity in the German Cancer Research Centre in Heidelberg, I have created a greedy search algorithm to maximise our likelihood score and detect RNA modifications from the signal data. To improve the accuracy, we included a posterior likelihood calculation into our score and I also coded a birth-death algorithm. Then, I tested both methods with multiple reads (signal lists) per mRNA type which proved to be more accurate and created interesting results!

#Statistics, #Biology, #Mathematics, #Optimisation, #Algorithms, #RNA, #Interdisciplinary, #Coding, #Data