Abstract

Small RNAs (sRNAs) are 20-30nt non coding RNAs that act as guides for the sequence specific regulatory mechanism known as RNA silencing. Recent developments in high throughput sequencing revealed a highly complex and diverse population of sRNAs. However, only a small proportion of the reads could be assigned to known classes of sRNAs such as microRNAs, trans acting siRNAs, heterochromatin RNAs.

 

The analysis of the un-annotated sRNA like reads in S. Lycopersicum samples led to the development of quality check procedures and machine learning techniques adapted to the characteristics of sRNA data. To evaluate the quality of the samples, the complexity and sample similarity indexes were proposed. Also, to diminish technical biases, several normalization procedures were investigated. The outcome of the QC consisted of expression profiles assigned to each sRNA. The step towards patterns was conducted using unsupervised learning (clustering with a non-standard correlation based  distance), simplified correlation based on patterns and a novel loci detection approach based on genome location, expression profile and size class distribution.

 

The analysis revealed novel characteristics of sRNAs in a developmental system and helped a better understanding of the mode of action of known classes of sRNAs, such as miRNAs.