RNA sequencing (RNAseq) has become the method of choice for transcriptome analysis, yet no consensus exists as to the most appropriate pipeline for its analysis, with current benchmarks suffering important limitations. Here, we address these challenges through a rich benchmarking resource harnessing (i) two RNAseq datasets including ERCC ExFold spike-ins; (ii) Nanostring measurements of a panel of 150 genes on the same samples; (iii) a set of internal, genetically-determined controls; (iv) a reanalysis of the SEQC dataset; and (v) a focus on relative quantification (i.e. across-samples). We use this resource to compare different approaches to each step of RNAseq analysis, from alignment to differential expression testing. We show that methods providing the best absolute quantification do not necessarily provide good relative quantification across samples, that count-based methods are superior for gene-level relative quantification, and that the new generation of pseudo-alignment-based software performs as well as established methods, at a fraction of the computing time. We also assess the impact of library type and size on quantification and differential expression analysis. Finally, we have created a R package and a web platform to enable the simple and streamlined application of this resource to the benchmarking of future methods.
Chronic kidney diseases (CKDs) are a global health concern, necessitating a comprehensive understanding of their complex pathophysiology. This study explores the use of 2 complementary multidimensional -omics data integration methods to elucidate mechanisms of CKD progression as a proof of concept. Baseline biosamples from 37 participants with CKD in the Clinical Phenotyping and Resource Biobank Core (C-PROBE) cohort with prospective longitudinal outcome data ascertained over 5 years were used to generate molecular profiles. Tissue transcriptomic, urine and plasma proteomic, and targeted urine metabolomic profiling were integrated using 2 orthogonal multi-omics data integration approaches, one unsupervised and the other supervised. Both integration methods identified 8 urinary proteins significantly associated with long-term outcomes, which were replicated in an adjusted survival model using 94 samples from an independent validation group in the same cohort. The 2 methods also identified 3 shared enriched pathways: the complement and coagulation cascades, cytokine–cytokine receptor interaction pathway, and the JAK/STAT signaling pathway. Use of different multiscalar data integration strategies on the same data enabled identification and prioritization of disease mechanisms associated with CKD progression. Approaches like this will be invaluable with the expansion of high-dimension data in kidney diseases.
A major challenge in sequencing-based spatial transcriptomics (ST) is resolution limitations. Tissue sections are divided into hundreds of thousands of spots, where each spot invariably contains a mixture of cell types. Methods have been developed to deconvolute the mixed transcriptional signal into its constituents. Although ST is becoming essential for drug discovery, especially in cardiometabolic diseases, to date, no deconvolution benchmark has been performed on these types of tissues and diseases. However, the three methods, Cell2location, RCTD, and spatialDWLS, have previously been shown to perform well in brain tissue and simulated data. Here, we compare these methods to assess the best performance when using human data from cardiovascular disease (CVD) and chronic kidney disease (CKD) from patients in different pathological states, evaluated using expert annotation. In this study, we found that all three methods performed comparably well in deconvoluting verifiable cell types, including smooth muscle cells and macrophages in vascular samples and podocytes in kidney samples. RCTD shows the best performance accuracy scores in CVD samples, while Cell2location, on average, achieved the highest performance across all test experiments. Although all three methods had similar accuracies, Cell2location needed less reference data to converge at the expense of higher computational intensity. Finally, we also report that RCTD has the fastest computational time and the simplest workflow, requiring fewer computational dependencies. In conclusion, we find that each method has particular advantages, and the optimal choice depends on the use case.