Illumina, PacBio, and Oxford Nanopore Technologies (ONT) are now among the most widely used next-generation sequencing (NGS) platforms in the world. Each one has their own strengths and limitations, and which one to use depends on the needs of your project. Below, I share some general comparisons that may help you decide.
Cost In general, Illumina remains the most cost-effective platform for short-read sequencing. Per Gb cost can often be below $50 now, especially on NovaSeq or NextSeq series. PacBio and ONT used to be more expensive, with older platforms costing around $1000-2000 per Gb. But things are changing. PacBio's newer Revio system and ONT's PromethION flowcells now allow higher throughput, which brings the cost down, especially for whole-genome or large transcriptome projects. Still, total cost also depends on experimental design, read depth, and coverage requirement.
Error Rates Illumina has long been known for low error rate, around 0.1-0.5%. This is enough for most variant calling and gene expression studies. For long-read platforms, error rate is higher - traditionally 10-15% for both PacBio and ONT. But newer PacBio CCS (circular consensus sequencing) protocol produces so-called HiFi reads with accuracy close to Illumina. It works by sequencing the same molecule multiple times and building a consensus, but it does increase cost and reduce total output.
ONT has also made some progress: Q20+ chemistry and duplex reads help reduce error rate - but still not as accurate as HiFi or Illumina. Some tools now can polish ONT reads with help of reference or short reads, but this adds complexity.
Read Lengths This is where PacBio and ONT shine. Illumina normally produces 2x150bp or 2x300bp reads, which is enough for most gene-level analysis but limited for structural variants or assembly. PacBio can routinely generate 15-25kb HiFi reads. ONT can produce even longer reads - 100kb or longer - with ultra-long protocols. In some special projects (e.g. telomere-to-telomere or repeat-rich regions), ONT long reads have clear advantage.
Applications Illumina has been the standard for most classical applications like whole exome sequencing (WES), RNA-seq, ChIP-seq, and Hi-C. Many existing workflows and pipelines were developed based on short reads. If your study fits into these well-established frameworks, Illumina is still very suitable and reliable.
But for projects requiring full-length transcript isoforms, de novo assembly, large indel detection, or epigenetic modifications (e.g. methylation), then long-read platforms like PacBio and ONT become better choices. Also, in microbiome and metagenomics studies, long reads help to resolve species or even strain level structure, especially when reference is incomplete.
Bioinformatics Analysis Illumina data is easier to work with. Many tools are mature and stable. Common mappers like BWA or STAR, and variant callers like GATK, are all well-supported. PacBio and ONT data are more challenging at first, due to higher error and more variability.
But now, many tools like minimap2, Flye, Shasta, and NextDenovo support long-read analysis. PacBio HiFi data can be processed by tools like Hifiasm, and ONT assemblies can be polished by Medaka or Homopolish. ONT basecalling (e.g. Guppy) is still being updated frequently - sometimes too frequent, so it is a bit hard to keep stable pipeline.
One challenge for long-read bioinformatics is that pipeline design depends more on project goal. Whether to do assembly first or align to reference? Should we do error correction before variant calling? These questions don't always have one fixed answer. That's why it is important to work with people experienced with these platforms, otherwise mistakes can occur and waste time.
Other Considerations Access to sequencing core or service provider is still important. Illumina instruments are widely available - nearly every core has it. PacBio and ONT, especially high-throughput models, are less common but growing.
Also, one should consider platform maturity. Illumina is more stable - not changing much anymore. PacBio and ONT are still developing fast. Sometimes their chemistry or basecalling software changes suddenly, and users must adjust. ONT often updates nanopore proteins and models, which is exciting but not always compatible with older tools. PacBio system is also quite closed - basecaller is proprietary - so some flexibility is missing.
We also observe that long-read data can be harder to revisit after a few years if formats or tools change. So, planning ahead is useful.
About the Author: Justin T. Li received Ph.D. in Neurobiology from University of Wisconsin-Madison in 2000, and M.S. in Computer Science from University of Houston in 2001. He was Assistant Professor in Medical School at University of Minnesota from 2004 to 2009, and Chief Bioinformatics Officer at LC Sciences (Houston) from 2009 to 2013. In 2013 he started AccuraScience, focusing on bioinformatics analysis for clients in research and biotech. Dr. Li has published over 40 peer-reviewed papers in bioinformatics and computational biology. More info at: https://www.accurascience.com/our_team.html.
Need assistance in your Illumina, PacBio or Oxford Nanopore data analysis project? Send us an inquiry, chat with us online, or reach us in other ways!