8 Future perspectives
This chapter gives potential improvements in context of the three sub-goals of our research. Many further perspectives are covered in the papers, therefore, this section covers only additional perspectives that are not discussed in the papers.
Firstly, for miRNA target prediction, it is important to improve the computational speed of our two-step SVM model at the classification phase. A current computational limitation of our model at classification is mainly caused due to its usage of non-linear kernel at the second or global level. Therefore, optimizing the second level classifier with a linear kernel is a simple solution to improve the computational speed at the classification phase. Moreover, training a SVM model with AGO pull-down data is simple, but it can be a very effective approach. However, existing AGO pull-down experiments provide no information regarding difference of miRNA binding sites between controls and transfected samples, therefore some scores, such as log-ratio values in microarray, need to be calculated for SVM training.
Secondly, the results from our study in miRNA high-throughput experiments can potentially improve other studies in both miRNA target prediction and miRNA and other ncRNAs because they mainly rely on the data from microarray, proteomics, and next generation sequencing for their model building and statistical analysis. Further interesting work can be an expansion of the analysis with different types of data, such as expression profiles of NCI-60 microarray data sets or time points data.
Thirdly, understanding of ncRNAs characteristics and their interactions with miRNAs becomes more important because ncRNA:ncRNA interactions are likely involved in many gene regulations. In addition to cis-NATs and CARs, it is interesting to expand the same approach to other types of ncRNAs, such as long ncRNAs.
In conclusion, using data from next generation sequencing as well as considering the results from miRNA high-throughput experiments is most likely to enhance other studies in both miRNA target prediction and miRNA and other ncRNAs.