Comparison of Stranded and Non-stranded RNA-Seq in Predicting Small RNAs in a Non-model Bacterium

Abstract

Thanks to their diversity, non-model bacteria represent an inexhaustible resource for microbial biotechnology. Their utilization is only limited by our lack of knowledge regarding the regulation of processes they are capable to perform. The problem lies in non-coding regulators, for example small RNAs, that are not so widely studied as coding genes. One possibility to overcome this hurdle is to use standard RNA-Seq data, gathered primarily to study gene expression, for the prediction of non-coding elements. Although computational tools to perform this task already exist, they require the utilization of stranded RNA-Seq data that must not be available for non-model organisms. Here, we showed that trans-encoded small RNAs can be predicted from non-stranded data with comparable sensitivity to stranded data. We used two RNA-Seq datasets of non-type strain Clostridium beijerinckii NRRL B-598, which is a promising hydrogen and butanol producer, and obtained comparable results for stranded and non-stranded datasets. Nevertheless, the non-stranded approach suffered from lower precision. Thus, the results must be interpreted with caution. In general, more benchmarking for tools performing direct prediction of small RNAs from standard RNA-Seq data is needed so these techniques could be adopted for automatic detection.Thanks to their diversity, non-model bacteria represent an inexhaustible resource for microbial biotechnology. Their utilization is only limited by our lack of knowledge regarding the regulation of processes they are capable to perform. The problem lies in non-coding regulators, for example small RNAs, that are not so widely studied as coding genes. One possibility to overcome this hurdle is to use standard RNA-Seq data, gathered primarily to study gene expression, for the prediction of non-coding elements. Although computational tools to perform this task already exist, they require the utilization of stranded RNA-Seq data that must not be available for non-model organisms. Here, we showed that trans-encoded small RNAs can be predicted from non-stranded data with comparable sensitivity to stranded data. We used two RNA-Seq datasets of non-type strain Clostridium beijerinckii NRRL B-598, which is a promising hydrogen and butanol producer, and obtained comparable results for stranded and non-stranded datasets. Nevertheless, the non-stranded approach suffered from lower precision. Thus, the results must be interpreted with caution. In general, more benchmarking for tools performing direct prediction of small RNAs from standard RNA-Seq data is needed so these techniques could be adopted for automatic detection.