Dealing with Experimental Noise
The applicability of CLIP-derived procedures is contingent to the extent of their background. A general assumption of these protocols is that random binding is hampered by the harsh washing steps succeeding immunoprecipitation. Moreover, electrophoretic resolution of the RNPs under study ensures an additional enrichment step with respect to adventitiously co-partitioned RBPs. Preferred control samples have normally been acquired from cells where the gene encoding the RBP of interest is knocked out, untagged or which have not been UV treated (Fig 2B-C) 22,23,27,37. In principle, these samples are not expected to produce any noticeable signal upon SDS-PAGE visualisation. Consequently, libraries emerging from the sequences in the areas corresponding to the expected migration of the RNP under study should contain 100-fold less unique cDNAs than those resulting from fully treated samples 37. Nonetheless, meta-analysis of 30 published CLIP datasets identified widespread and replicable background reads across most of them 38. To quantify technical background, prior work sequenced libraries proceeding from RNA-protein adducts that did not migrate with the duplexes of interest (Fig. 2C)38. Adjusting output signals to those originated by background characterisation was shown to dramatically improve the identification of RNA recognition sequences 38. Based on relevant control samples, background correction can indeed statistically disregard binding events resulting from non-specific transcript cross-linking to incidentally proximal RBPs of interest, co-purified RNPs and RNA aptamers adopting an epitope-mimicking conformation.
A number of CLIP-related approaches have been developed that have been specifically designed to further improve signal-to-noise ratios. These methods take full advantage of the fact that cross-lined RNPs can withstand highly denaturing purification conditions. These techniques include cross-linking and analysis of cDNAs (CRAC), ultraviolet crosslinking and affinity purification (uvCLAP) and denaturing CLIP (dCLIP). All of these protocols employ (tandem) affinity-based purification to recognise specific epitope tags that are compatible with denaturing purification conditions and/or strong detergents (e.g., HIS-tag and streptavidin-binding peptides sequences; Fig. 1)23,39,40. Admittedly, epitope tagging can alter the stability, expression, and RNA-binding capacity of the RBP of interest. Advantageously, however, specific tag recognition-based partition spares the need for high-quality antibodies, which are not readily available for less well studied model organisms. Accordingly, these approaches can considerably reduce background reads in resulting cDNA libraries. This property also makes them good alternatives to inspect scarcer RNPs, namely yeast pre-ribosomal complexes or those encompassing non-professional RBPs 2,23.