Table 3.

Comparison of long-read and short-read sequencing technologies.

	Illumina	10× linked-read	Short-read RNA-seq
Avg read length	2× 150–250 bp	IL platform dependent	2× 75 bp
Max read length	2× 250 bp	(~100 kb span)	2× 100 bp
Accuracy per nucleotide1	>99%	(see Illumina)	(see Illumina)
Error bias	Substitutions in high/low GC regions	(see Illumina)	(see Illumina)
Coverage bias	Low coverage of high/low GC regions. Mapping issues with highly homologous regions	(see Illumina)	Illumina biases and additional: ligation bias due to the reverse transcriptase enzyme101, protocol differences poly-A only versus ribodepletion
Accuracy after error correction2	N/A	N/A	N/A
Sample requirements PCR-free3	1–2 μg/(default > 500 ng)	IL + controller chip 1 ng	0.1–1 μg
Low-throughput sample requirements	100 ng3	–	25 ng
Base modifications	Bisulfite sequencing required	–	–

	Nanopore	PacBio	BNG	Hi–C
Avg read length	15–20 kb	10–15 kb	~100 kb resolution, variants >500 bp	~1 kb–1 Mb resolution102
Max read length	>800 kb103,104	>60 kb103–105
Accuracy per nucleotide1	60–85%103,104,106	>85%105	no base pair resolution	no base pair resolution
Error bias	Small indels, mostly deletions49,107	Small indels, mostly insertions49,107	N/A	N/A
Coverage bias	Truncation of homopolymers and low-complexity regions103	Homopolymers	Fragile sites, nick enzyme-dependent108	Biases depend on protocol used: restriction enzymes, PCR, and IL seq102,109
Accuracy after error correction2	After 1D2 97% After hybrid correction: >99%	After CCS: 95–99%60.	N/A	N/A
Sample requirements3	1 μg–400 ng HMW 0.4–1 μg HMW	10 μg HMW	0.3–0.9 μg HMW	1–10 million cells102,109
Low-throughput sample requirements	10–100 ng	400–800 ng	300 ng	1000 cells110
Base modifications	Theoretically all	Methylation, Mostly bacterial	–	–

Comparison of Pacific Biosciences (PacBio), Oxford Nanopore Technologies (ONT), Illumina (IL), 10× Genomics linked-read sequencing on the Illumina platform (10×), RNA sequencing on the Illumina platform (RNA-seq), BioNano Genomics (BNG) and the genome-wide chromatin conformation capture protocol Hi–C (Hi–C). Many characteristics of 10× and RNA-seq are shared with IL, since the same sequencing platform is used. Also note that Hi–C protocols are under active development, they vary in biases, sample requirements and use of IL sequencing^102,109,110. Unreferenced values are derived from the manufacturer’s websites last accessed in October 2020 [Oxford Nanopore Technologies (https://nanoporetech.com/), PacBio (https://www.pacb.com/), Illumina (https://www.illumina.com/), 10x Genomics (https://www.10xgenomics.com/), Illumina Stranded mRNA Prep (https://www.illumina.com/products/by-type/sequencing-kits/library-prep-kits/stranded-mrna-prep.html), Bionano Genomics (https://bionanogenomics.com/)].

¹Reported accuracy for PacBio and ONT strongly depends on the sequencing platform version and polishing steps. Using regular single-pass sequencing without self-correction, both PacBio and ONT theoretically have per-nucleotide error rates of ~15%^103–105 but previous versions of the MinION up to ~40%¹⁰⁶.

²The latest ONT and PacBio technologies attain >99% accuracy for de novo human assemblies. PacBio achieves >99.8% accuracy using circular consensus sequencing (CSS) where the same read is sequenced many times and averaged, although this limits read length to ~13 kb⁶⁰. ONT reports >99% after polishing with short reads (hybrid correction) which is necessary due to truncation of homopolymers and low-complexity regions¹⁰³. ONT 1D² technology sequences both DNA strands and uses consensus to attain >97% whilst maintaining read lengths, although only ~60% of the molecules can be sequenced using this approach [Oxford Nanopore Technologies (https://nanoporetech.com/)].

³Sample requirements as listed by the manufacturer and dependent on the library preparation method used, e.g., insert size and use of PCR, as well as the exact version of the machine. High molecular weight (HMW) DNA is required to attain long read lengths, but the read length of PacBio is limited by the polymerase and for ONT by the length of the DNA molecules hence it can report ultra-long reads >800 kb¹⁰³. The minimum sample amount of 10 ng listed by ONT is likely insufficient for a human genome. Whilst for IL in practice smaller amounts e.g., 50 ng are used as low-throughput minimum.