Next-generation sequencing (NGS) is a powerful platform that enables the profiling of known sequences and also the discovery of novel small noncoding RNAs with unprecedented sensitivity (Creighton et al. 2009; Metzker 2010; Tam et al. 2014). Differential expression of miRNAs and importantly detection of novel miRNAs or sequence variations (isomiRs) can be obtained (Ozsolak and Milos 2011). Initially, adapters specifically modified to target miRNAs and other small RNAs that have a 3' hydroxyl group are ligated to each end of the RNA molecules, and an RT reaction is used to generate single-stranded cDNA that is then PCR amplified. Sequencing data of libraries generated by RNA ligases revealed secondary preferences of these enzymes. A strong bias towards certain small RNAs has been reported thereby preventing determination of absolute numbers of small RNAs (Linsen et al. 2009). This bias was systematic and highly reproducible and was shown to be independent of the sequencing platform but strongly determined by the method of small RNA library preparation (Tang et al. 2013). The technical error in small libraries can be reduced with the introduction of primers containing index sequences during the PCR amplification of the cDNA. Gel purification to prepare a library product for subsequent cluster generation will follow (Sorefan et al. 2012). Recently, improved protocols that require lower RNA input were developed. This is particularly important for circulating miRNAs as they display low abundance. Nevertheless, in most cases, 2-3 ml of plasma or serum must be processed for each sample, limiting the use of this method to the discovery phase in a limited number of pooled samples (Li et al. 2012; Nielsen et al. 2012), rather than applying it routinely in clinical cohorts. Further improvements may involve combination of microfluidic technology (Streets et al. 2014) and high-throughput sequencing and could offer more feasible solutions for larger studies. Recently, NGS workflows designed and optimised specifically for serum and plasma samples that utilise a smaller volume of input material, as little as 500 ql, became commercially available. However, one should exercise caution as these platforms have not been extensively tested. In addition, the overall cost is still substantially higher compared to other platforms, and bioinformatics support for the computational analysis and interpretation is required. Normalisation of the sequencing data can be challenging as a highly abundant miRNA in a sequencing sample may affect the number of sequencing reads that are available for detection and quantification for other miRNAs in the sample. The abundance of a given target can be represented as a percent of total reads, provided that the distribution of miRNA reads does not change substantially. The addition of a spike-in synthetic control during library preparation is an alternative approach. A good correlation between the input material and output reads, a reduction in systematic error and improved accuracy of base quality scores has been demonstrated (Jiang et al. 2011; Zook et al. 2012).
Inter-assay studies demonstrated that the various miRNA profiling technologies display significant differences in specificity, sensitivity, dynamic range and accuracy. On several occasions, poor correlation was observed between the diverse platforms suggesting that the interpretation of population studies performed using different technologies would be hampered by these technical limitations (Ach et al. 2008; Git et al. 2010; Jensen et al. 2011; Mestdagh et al. 2014; Tam et al. 2014). Hence, there is an apparent need for advances in technology and bioinformatics to establish a ‘universal standard’ that can accurately monitor the miRNA profile in the circulation and track expression differences with high sensitivity.