Robust Transcriptional Signatures For Low-input RNA Samples Based On Relative Expression Orderings

For high-throughput gene expression profiling,it is often difficult to obtain sufficient quantities of RNA from tissue biospies,laser-capture microdissection tissues,rare cell populations,clinical samples available only from cadavers formalin-fixed paraffin-embedded(FFPE)tissues,or rare cell populations.In such cases the preamplification of RNA is necessary prior to quantify gene expression levels.Although many low-input RNA amplification techniques have been developed to address this issue,they usually bring bias with high levels of technical noise due to inherent defects in amplification principles and batch effects.Alongside the technological breakthroughs that have facilitated the large-scale generation of low-input RNA amplification samples transcriptomic data,it is important to consider the specific computational and analytical challenges that still have to be overcome.Herein,low-input RNA amplification samples were beginning with 1000 pg,100pg,50 pg 25pgamplified and profiled by Smart-seq,DP-seq and CEL-seq amplification techniques,then,Illumina Hi Seq 2000 platform was used to quantify gene expression levels.The other group of low-input RNA amplification samples were beginning with 1000 pg,250pg,50 pg 10pg amplified and profiled by the Whole-Genome DASL technique,then,the Illumina Human Ref-8 v3.0 expression beadchip microarray platform was used to quantify gene expression levels.In this study,through respectively comparing gene expression profiles between low-input RNA amplification samples and their paired high-input RNA amplification samples,we firstly showed that the expression measurements of thousands of genes had at least two-fold change in low-input RNA amplification samples compared with paired high-input amplification RNA samples.Therefore,for a transcriptional signature based on risk scores summarized from the expression levels of the signature genes,the risk score thresholds determined from RNA high-input amplification samples could not be applied to low-input RNA amplification samples,and vice versa.On the other hand,we showed that around 90% relative expression orderings(REOs)of gene pairs in the high-input amplification RNA samples were maintained in their paired 1ng,100 pg,50pg low-input RNA amplification samples,amplified by Smart-seq,DP-seq and CEL-seq,measured in the Illumina Hi Seq 2000;and at least 90% relative expression orderings(REOs)of gene pairs in the high-input amplification RNA samples were maintained in their paired 1ng,250 pg low-input amplification RNA samples amplified by the PCR-amplified RNA amplification technique in the Illumina Human Ref-8 v3.0 expression beadchip microarray platform;These result suggested that the REOs of gene pairs were highly robust against amplification bias even with as low as 250 pg amplification samples measured by the Illumina Human Ref-8 v3.0 expression beadchip microarray platform and even with as low as 25 pg RNA amplification RNA samples amplified by Smart-seq,DP-seq measured in the Illumina Hi Seq 2000.Finally,as a case study,we developed a REOs-based signature from high-input RNA amplification breast cancer and lymphoma tissues,and validated that it could successfully distinguish Raji cells from MCF-7 cells using low-input RNA amplification samples.In conclusion,the high-input amplification RNA samples can be fully exploited to identify REOs-based classification signatures which could be robustly applicable to low-input RNA amplification samples.Our analysis allows biologists to gain the reliable and reproducible results and select the most suitable amplification methods for low-input RNA amplification samples.