Conventional single-cell RNA sequencing (scRNA-seq) methods are limited to gene-level expression profiling, overlooking the fact that many genes produce multiple isoforms, which can encode functionally distinct proteins. To address this, we have co-developed long-read scRNA-seq to enable cell-specific, isoform-level analyses. However, given the novelty of long-read single-cell approaches, their limitations and sensitivities remain largely unknown, and guidelines for optimal cell counts and sequencing depths are lacking. To address this gap, we conducted benchmarking analyses using matched long-read bulk and scRNA-seq data.
In our pilot study, we used carcinoma samples to evaluate how sequencing depth impacts the detection of genes and isoforms in long-read scRNA-seq. We generated libraries with read depths ranging between 500 to 24,000 median reads per cell. Overall, we find that isoform and gene detection sensitivity are similar to bulk long-read sequencing. Analysing the agreement in the isoforms identified between the techniques, we discovered the highest overlap (56%) at 24,000 median reads per cell (~11.5 million total reads corresponding to 473 cells). Across all read depths trialled, single-cell sequencing identified a larger number of unique isoforms, though the relative proportion of this category decreased as sequencing depth increased. These results indicate that false positive isoforms are likely an underappreciated challenge with long-read scRNA-seq and that robust isoform discovery requires both greater isoform filtering and sequencing depths per cell than previously assumed.
To validate our long-read scRNA-seq method in a biological context, we analysed organoids modelling prenatal neurodevelopment – a process characterised by complex isoform regulation. Using a deeply sequenced library, we detected known cell-type-specific RNA regulatory events, including the switch from PTBP1 to PTBP2 during neuronal differentiation. These results demonstrate the potential of long-read scRNA-seq to capture cell-type-specific isoform and gene expression in biological contexts. Importantly, our benchmarking study provides valuable insights for future long-read scRNA-seq experimental design.