A bias-reducing strategy in profiling small RNAs using Solexa

Abstract

Small RNAs (smRNAs) encompass several different classes of short noncoding RNAs. Progress in smRNA research and applications has coincided with the advance of techniques to detect them. Next-generation sequencing technologies are becoming the preferred smRNA profiling method because of their high-throughput capacity and digitized results. In our small RNA profiling study using Solexa, we observed serious biases introduced by the 5' adaptors in small RNA species coverage and abundance; therefore, the results cannot reveal the accurate composition of the small RNAome. We found that the profiling results can be significantly optimized by using an index pool of 64 customized 5' adaptors. This pool of 64 adaptors can be further reduced to four smaller index pools, each containing 16 adaptors, to minimize profiling bias and facilitate multiplexing. It is plausible that this type of bias exists in other deep-sequencing technologies, and adaptor pooling could be an easy work-around solution to reveal the "true" small RNAome.

FIGURE 1.

(A) A diagram of Illumina v1.5 protocol using customized 5′ adaptors. (B) Heatmap of clustered 64 customized 5′-adaptor results (partial shown; see Supplemental Material for the whole heatmap). Total reads of each adaptor from Test 1 were normalized to the max of the sum of all miRNA reads in each index by scaling up total reads in each adaptor to this max. Each miRNA in an adaptor was scaled up according to the ratio of the total reads in an index to the max. Then, the normalized reads were log₂-transformed and clustered as complete linkage and used to generate the heatmap. (C) Result comparison of using customized 5′ adaptor versus Illumina default 5′ adaptor. The top 20 miRNAs are shown.

FIGURE 2.

(A) Heatmaps of 8 × 8 pools versus 16 × 4 pools versus 32 × 2 pools in Test 1. (B) Pearson correlation: Designed 8 × 8 pool versus randomized 8 × 8 pools in Test 1. (C) Heatmap of lane comparison in Test 2. (Lanes 2,6) Duplicate run of 8 × 8 pools; (lane 3) 16 × 4 pools; (lane 4) 32 × 2 pools; (lane 5) 64 × 1 pool. Total reads of miRNAs in different lanes were normalized and log₂-transformed to generate the heatmap.

FIGURE 3.

Validating pool-adaptor results by Northern blotting analysis. (A) Lane-to-lane read count comparison of miRNAs used for Northern blotting analysis: miR-21, miR-1, miR-126, miR-205, miR-101 from human pool RNA samples that were sequenced with 8 × 8 adaptor pools, default 5′ adaptor, and breast RNA that was sequenced using 8 × 8 adaptor pools. (B) Northern blotting analyzes of miR-21, miR-1, miR-126, miR-205, and miR-101; U6 snRNA was used as RNA loading control. (Lane 0) RNA decade marker; (lane 1) human RNA pool; (lane 2) breast RNA.