Oxford Nanopore MinION Direct RNA-Seq for Systems Biology

Abstract

Long-read direct RNA sequencing developed by Oxford Nanopore Technologies (ONT) is quickly gaining popularity for transcriptome studies, while fast turnaround time and low cost make it an attractive instrument for clinical applications. There is a growing interest to utilize transcriptome data to unravel activated biological processes responsible for disease progression and response to therapies. This trend is of particular interest for precision medicine which aims at single-patient analysis. Here we evaluated whether gene abundances measured by MinION direct RNA sequencing are suited to produce robust estimates of pathway activation for single sample scoring methods. We performed multiple RNA-seq analyses for a single sample that originated from the HepG2 cell line, namely five ONT replicates, and three replicates using Illumina NovaSeq. Two pathway scoring methods were employed-ssGSEA and singscore. We estimated the ONT performance in terms of detected protein-coding genes and average pairwise correlation between pathway activation scores using an exhaustive computational scheme for all combinations of replicates. In brief, we found that at least two ONT replicates are required to obtain reproducible pathway scores for both algorithms. We hope that our findings may be of interest to researchers planning their ONT direct RNA-seq experiments.

Keywords: HepG2; MinION; RNA-seq; nanopore technology; pathway activation; systems biology; transcriptomics.

Figure 1

The overall workflow of the experiment. RNA was extracted from the HepG2 cell line and technical replicates were performed using Illumina NovaSeq 6000 (3 replicates) and Oxford Nanopore MinION (5 replicates). Expression of protein-coding transcripts was quantified using Salmon.

Figure 2

Comparison of intra- and inter-platform variation of gene quantification for Illumina and ONT. Data were restricted to genes that were observed in all replicates for each platform. Gene expression was measured in TPM units. (a) Distribution of coefficient of variation for both platforms. (b) Between replicates pairwise gene correlation matrix with Spearman coefficients for ONT platform. (c) Between replicates pairwise gene correlation matrix with Spearman coefficients for Illumina platform. (d) Correlation between average gene expression ranks for Illumina and ONT platform. Genes are represented by dots. Horizontal and vertical spurs indicate genes not detected by ONT (n = 1850) and Illumina (n = 554) platforms, respectively.

Figure 3

Total number of detected features for various number of replicates. Different combinations of experimental replicates ONT-1, …, ONT-5 were generated. Each transcript (a) or gene (b) was claimed to be expressed if it was detected in at least one replicate included in the combination.

Figure 4

Pairwise correlation between pathway scores. Different combinations of experimental replicates ONT-1, …, ONT-5 and ILMN-1, …, ILMN-3 were generated. For every two combinations, pathway activation scores were estimated. Distribution of Spearman correlation coefficients between pathway scores using either singscore (a) or ssGSEA (b) algorithms was plotted for each number of replicates.