Validation of methods for Low-volume RNA-seq

Recently, a number of protocols extending RNA-sequencing to the single-cell regime have been published. However, we were concerned that the additional steps to deal with such minute quantities of input sample would introduce serious biases that would make analysis of the data using existing approaches invalid. In this study, we performed a critical evaluation of several of these low-volume RNA-seq protocols, and found that they performed slightly less well in metrics of interest to us than a more standard protocol, but with at least two orders of magnitude less sample required. We also explored a simple modification to one of these protocols that, for many samples, reduced the cost of library preparation to approximately $20/sample.

: Total TruSeq cDNA library yields made with a given amount of input total RNA. Yields measured by Nanodrop of cDNA libraries resuspended in 25µL of EB. The italicized samples were unusually low, and when analyzed with a Bioanalyzer, showed abnormal size distribution of cDNA fragments. We considered the two libraries with lower than usual concentration to be Thus, we consider 70 ng of total RNA to be the conservative lower limit to 85 the protocol. While this is about 30% smaller than the manufacturer suggests, it 86 is still several orders of magnitude larger than we needed it to be. We therefore 87 considered using other small-volume and "single-cell" RNA-seq kits, which we 88 had less experience with and less faith in the data. to the lowest number of mapped reads in any sample in order to provide a fair 117 comparison between protocols. 118 We were interested in the fold-change of each D. virilis gene across the four 119 samples, rather than the absolute abundance of any particular gene. Therefore, (1) 125 We filtered the D. virilis genes for those with at least 20 mapped fragments 126 in the sample with 20% D. virilis, then calculated an independent linear re-127 gression for each of those genes. As expected, for every protocol, the mean 128 slope was 1 (t-test, p < 5 × 10 −7 for all protocols). Similarly, the average in-129 tercepts for all protocols was 0 (t-test, p < 5 × 10 −7 for all protocols). Also  Although the SMART-seq2 was the cheapest of the protocols, we wondered 149 whether it could be performed even more cheaply without compromising data 150 quality. This would enable us to include more biological replicates in the future 151 experiments for which we are evaluating these protocols. In the original protocol,

152
we noticed that roughly 60% of the cost came from the Nextera XT reagents.

153
Thus, reducing the cost of tagmentation was the obvious goal to target. 154 We made additional libraries, again starting with 1ng of total RNA. We  Table 4). This is despite the additional cycles of enrichment, which 165 improved yield.

166
Because we used a common set of pre-amplified cDNA samples that was 167 performed in a distinct pre-amplification from experiment 2, we can estimate 168 the contribution of that pre-amplification to the overall variation. If, in fact, the 169 pre-amplification is a major contributor to the variation, then we would expect 170 to find that the correlation between, for instance, the slopes of two runs of the   than the correlation between the slopes of two runs using the same pre-amplified 173 cDNA pools.

174
Unsurprisingly, the sets of samples that used the same preamplification were 175 more correlated with each other than with the set of samples that used a separate 176 pre-amplification (Fig. 3). By analogy to dual-reporter expression studies[6], we 177 term variation along the diagonal "extrinsic noise" (η ext = std(m 1 + m 2 )), and 178 variation perpendicular to the diagonal "intrinsic noise" (η int = std(m 1 − m 2 )), 179 being intrinsic to the pre-amplification step. Using that metric, the intrinsic  One of the more striking results is that costs can be significantly reduced by 219 simply performing smaller reactions, without noticeably degrading data quality. 220 We do not suspect this will be true for arbitrarily small samples, such as from