AbstractThe attachment of unique molecular identifiers (UMIs) to RNA molecules prior to PCR amplification and sequencing, makes it possible to amplify libraries to a level that is sufficient to identify rare molecules, whilst simultaneously eliminating PCR bias through the identification of duplicated reads. Accurate de-duplication is dependent upon a sufficiently complex pool of UMIs to allow unique labelling. In applications dealing with complex libraries, such as total RNA-seq, only a limited variety of UMIs are required as the variation in molecules to be sequenced is enormous. However, when sequencing a less complex library, such as small RNAs for which there is a more limited range of possible sequences, we find increased variation in UMIs are required, even beyond that provided in a commercial kit specifically designed for the preparation of small RNA libraries for sequencing. We show that a pool of UMIs randomly varying across eight nucleotides is not of sufficient depth to uniquely tag the microRNAs to be sequenced. This results in over de-duplication of reads and the marked under-estimation of expression of the more abundant microRNAs. Whilst still arguing for the utility of UMIs, this work demonstrates the importance of their considered design to avoid errors in the estimation of gene expression in libraries derived from select regions of the transcriptome or small genomes.