The Sequencing Revolution: enabling personal genomics and personalised medicine

Figure 1 Advances in the development of sequencing technologies have resulted in an increase in data output with a dramatic decrease in cost. This graph compares calculated sequencing costs for one complete haploid human genome sequence (23 chromosomes, three billion bases) * estimated from literature ** marketing figures

It has been 10 years since the completion of the first draft of the human genome. Today, we are in the midst of a full assault on the human genetic code, racing to uncover the genetic mechanisms that affect disease, aging, happiness, violence … and just about every imaginable human variation. Advances in DNA sequencing technology have enabled individuals to have their own genomes sequenced rapidly, cheaply and in astonishing detail. The sequencing revolution is also changing the way the pharmaceutical industry develops, tests and targets new medicines.

Although humans have recognised hereditary traits in plants, animals and humans for thousands of years, it was only at the beginning of the 20th Century that the concept of the gene arose, when researchers working on fruit flies narrowed down the material basis of trans – missible traits to chromosomes1. Moreover, it was not until 1953, when Watson and Crick determined the double helix structure of DNA, that nucleic acids and their components became the central focus of genetic research2.

Beginning in the 1960s, the first nucleic acid sequencing attempts used chromatography. While this was neither reliable nor robust enough for routine use, it resulted, in 1972, in the publication of the first complete nucleotide sequence of a gene – from the MS2 virus3. Incredibly, it took only four more years to publish the entire 3,600-base sequence of the MS2 genome. Then, in 1977, Frederik Sanger, at the University of Cambridge, published the chain terminator method of DNA sequencing4 which became the method of choice for routine sequencing in academic centres. In the 1980s, the development of dye terminators (nucleotides with unique fluorescent markers) made it possible to automate Sanger’s method, speeding up the process tremendously. By 2000, when the first draft of the human genome (the 24 chromosome haploid sequence), with three billion base pairs, was sequenced, the project had taken only a decade, but had required hundreds of sequencing machines, employed teams of researchers around the globe, and cost three billion dollars. Part of the reason the sequence was completed so quickly was the race between the public consortium (a worldwide network of research centres) and private efforts led by the Celera Corporation to be the first to sequence and profit from the genome’s information5,6.

During the race to complete the human genome, research centres made technical advances on fluorescence based DNA sequencing – such as pyrosequencing, reversible nucleotide terminators and synthesis of ‘DNA colonies’, which allowed for the commercial realisation of Next Generation Sequencing (NGS).

The first NGS system, unveiled in 2004, was the 454 sequencing platform. This is based on a massively parallel pyrosequencing process, which sequences each DNA strand by synthesising a complementary strand7. For each testing cycle, a fluorescently tagged nucleotide (A, T, C or G) is added to the reaction. Each time the tagged nucleotide is incorporated, it generates a light signal of a specific strength, meaning the strength of the overall signal from a sample is proportional to the number of nucleotides incorporated in it. This value is converted into a nucleotide read. The latest 454 platform, the Genome Sequencer FLX, can sequence 500 million bases per 10 hour run at a cost of USD 7,500. At this rate, an entire human genome can be sequenced in one week at a cost of under USD 50,000.

Several other companies have emerged with platforms that can output more sequence data per run with even lower costs – the strongest competitors in the field include Illumina Solexa8, Applied Biosystems (AB) SOLiD9, Helicos Biosciences10, and Complete Genomics11. Where the Sequencer FLX reads strings averaging 660 bases in length, though, these systems are based on short (10 to 150 base) reads – with higher intrinsic error rates requiring repetitive sampling (or coverage) of the template DNA to build the final consensus sequence. The Illumina and Helicos platforms utilise variations on the same theme, but use reversible fluorescence based single nucleotide terminators for their synthesis. AB SOLiD and Complete Genomics use sequencing by ligation strategy with short fluorescently labelled DNA probes. Similar to the 454 platform, billions of parallel sequencing reactions are conducted and the machine run takes approximately one week. However, for these systems, the sequence cost for one complete human genome is under USD 10,000.

Figure 1 Advances in the development of sequencing technologies have resulted in an increase in data output with a dramatic decrease in cost. This graph compares calculated sequencing costs for one complete haploid human genome sequence (23 chromosomes, three billion bases) * estimated from literature ** marketing figures

Illumina’s current HiSeq platform can output 300 billion bases of reads in a single run. This volume of data output makes it possible to sequence several human diploid genomes on one machine with sufficient coverage to guarantee accuracy. Comparably, Applied Biosystems (part of Life Technologies Corporation) has announced the SOLiD 4 system, which produces 100 billion bases of reads with 99.9 per cent accuracy. This is achieved because SOLiD’s probe hybridisation technique sequences each base twice on the template. For both the HiSeq and the SOLiD 4, the cost for one human diploid sequence, with reliable accuracy, is approximately USD 5,000.

The Helicos platform boasts the only commercially available method for single molecule sequencing10. By eliminating steps used in other systems, it significantly reduces the biases and artefacts associated with sample preparation. It also allows the use of almost any nucleic acid as a starting material – including degraded DNA for applications in forensic research. Helicos has also published articles demonstrating direct RNA sequencing on their platform, raising the possibility of assessing the transcriptome in a single cell12.

Complete Genomics Inc. uses rolling circle template amplification, thereby creating a novel nanoball of DNA captured on a solid surface11. Because of the complexity of this system, the company has adopted a model of providing sequencing services rather than selling platforms. Complete Genomics aims to sequence 1000+ human individuals in 2010 and to expand data output by building satellite sequencing facilities.

Following rapid advancements in Second Generation sequencing, the first Third Generation systems will soon be available. Later this year, Pacific Biosciences13 and Life Technologies14,15 are both scheduled to unveil their single molecule sequencing platforms. Current technologies offer up to several hundred base read lengths and shorter processing time using large samples of genetic material, but Third Generation systems will achieve read lengths of greater than 1000 bases from a single DNA molecule, with less sample processing time and lower reagent costs. Long reads can fully span repetitive or structurally varied genomic regions.

Pacific Biosciences makes use of tiny compartments (called Zero Mode Waveguides) where the fluorescence signal of a single nucleotide can be recorded when it is briefly held in place by a DNA polymerase on the template strand (Figure 2). Life Technologies employs a FRET based system – the DNA polymerase is attached to an energy donor bead that can cause light emission from a fluorescently labelled nucleotide only when it is being incorporated in the DNA strand (Figure 2). As with Second Generation machines, the platforms employ optic systems to record emitted light which is converted to a nucleotide read. Both platforms add ‘redundant sequencing’– the DNA strand is sequenced twice to give the sequence 99.9 per cent accuracy. A third platform, Ion Torrent15 (acquired by Life Technologies), does not use optics, but semiconductor chips (Figure 2). The chips are fabricated to detect the hydrogen ion which is normally released when a polymerase incorporates a single nucleotide into the growing DNA strand. The result is a platform that is easier to manufacture and operate with lower costs and shorter run times. Ion Torrent promotes that their system will sell for USD 50,000 and can deliver sequence data, comparable to other platforms, in one hour15.

Figure 2 The third generation sequencing technologies. A) Pacific Biosciences system uses tiny wells (nanometres in diameter) called ‘Zero Mode Wave Guides’. The small volume allows detection of the light emission from a single nucleotide in the vicinity of the DNA polymerase as it is incorporated into the growing DNA strand. B) Life Technologies employs a five nanometre quantum dot nanocrystals (yellow) attached to the polymerase (white). As fluorescent nucleotides are incorporated into the DNA strand, a FRET [Fluorescence Resonance Energy Transfer signal (shown in green) is generated between the nanocrystals and the base, which is recorded by sensitive optics. C) Ion Torrent technology detects the presence of an H+ ion, which is released after a base is normally added to the DNA strand by a DNA polymerase (in green). The reaction occurs in nanolitre wells. The detection system (shown in panel D) employs an ion sensitive layer (green) built on top of a proprietary ion detector (beige) using semiconductor technology.”

Third Generation platforms promise advances in throughput, scalability, speed and cost over the current methods. However, nanopore based technologies, though further from commercial realisation, have the potential for even greater benefits. For instance, the DNA transistor16, currently under development by IBM Research, will offer true single molecule sequencing by decoding the electrical charge of nucleotides of DNA strands as they are threaded through a 1.5 to 3-nanometre-wide pore in a silicon chip. NABsys of Providence, Rhode Island, is testing a system called ‘Hybridisation-assisted Nanopore Sequencing Platform’17 which employs 6-base oligos hybridised to 100 kilobase DNA fragments. The single stranded DNA is threaded through a nanopore where the hybridised probes are detected by current changes. The software builds the entire sequence of the DNA fragment by deconvoluting the spacing of multiple independent 6-mer hybridisation experiments. NABsys plans to offer whole human genome sequencing for under USD 100. The technologies employing nanopore devices do not modify the native DNA strand and may be collectively referred to as Fourth Generation. The resulting processes are simpler and sequencing more accurate, with the potential to sequence the entire human genome in less than a day. The technical breakthrough in developing nanopore-based sequencing technology will be in controlling the motion of the DNA as it passes through the opening for accurate readings.

The question that still needs to be answered is: which platform should my company invest in or utilise for routine applications now? Because sample preparation and machine run times are similar, one useful principle is to choose platforms that suit current or foreseeable application needs. For de novo sequencing – especially for sequences with repeat regions, where long reads are usually desirable, the 454 platform offers the longest reads. For transcriptomics, chromosome immuno – precipitation and epigenetic studies, shorter reads are sufficient, and can be reliably addressed on Helicos, AB SOLiD and Illumina machines. (These recommendations will, of course, change when the advances in third generation sequencers are fully realised).

The reliability and data output standardisation of these sequencing platforms are currently the focus of a two year study led by the US Food and Drug Administration18. The Sequence Quality Control Initiative (SEQC) distributed identical samples to be sequenced by facilities worldwide employing different sequencing platforms. The data output was then distributed again to various centers using diverse analytical software. Due for publication at the end of 2010, the SEQC report will represent what may be the most extensive study of its kind, and will help develop better standardised protocols, especially for industries where sequence data is used for diagnostics.

The analysis of the first human genome draft naturally led to the question of the genetic relationship between hereditary differences and the onset of disease. In the last decade, various groups have mapped known variable regions among disease-linked and normal populations using microarray based methods in copy number variation (CNV) and genome wide association studies (GWAS)19. It is largely accepted that these studies have failed to adequately identify genetic disease markers, and that the real genetic links may lie not in common variants but in rare alleles. The resolution of individuals’ genomes and transcriptomes down to single nucleotides by sequencing may be necessary to isolate the alleles that explain the molecular basis of disease.

The next step, profiling at the genetic level, could advance medical care by identifying persons likely to benefit most from a particular medication and those most at risk of adverse reactions. Most drugs fail in Phase II of clinical trials, where the drug is tested on a small cohort of patients with the disease. Yet subgroups within the cohort who react positively or negatively to the drug may share genetic markers predictive of such responses. Isolating these markers would allow for the selection of the best responders for testing in the Phase III trials – resulting in better targeted, and therefore safer, more effective drugs. It is tempting to speculate that many compounds that failed in previous clinical trials can be rescued by correcting the patient stratification.

Clinical trials are the most costly part of drug development. Stratifying drug responders earlier in the process (i.e., the pre-clinical stage), will make clinical trials more efficient and less expensive. Most pharmacology studies employ animal models that unavoidably exhibit variable responses during pre-clinical testing. Some models may not be suitable because they differ unpredictably from humans. For example, primate models used for toxicology studies could be of mixed genetic backgrounds because of species inter-breeding, confounding the results of the study with a wide range of drug responses. Sequencing of each test animal will soon become a routine way to identify early markers of traits that influence a drug’s effects, giving an early window on the target population for testing in clinical trials.

Accompanying dramatic decreases in cost, increasing accessibility to sequencing tech – nologies has been a boon to evolutionary research. The full genomic sequences of 3,800 organisms, including 14 unique mammals, have already been deposited in public data archives. The number of human beings sequenced is also increasing: thus far, about 30 human genomes are available to the public in a database maintained by the Personal Genome Project (PGP), whose goals include the collection of 100,000 genomes to allow researchers to catalogue human genetic variation20.

Although few groups have the PGP’s capacity to maintain sequencing data at the rate that it is being produced, several large scale efforts have been founded to sequence smaller human and human-related samples. To name just a few, the Human Microbiome Project aims to sequence all microbial organisms in our bodies21, the Cancer Genome Atlas will map cancer cell lines and primary tumours22, and the Encode/modEncode project will attempt to catalogue all epigenetic markers and DNA binding protein sites in humans and selected model organisms23. Handling the flood of data remains the next biggest challenge in the field. The National Institutes of Health (NIH) and European Bioinformatics Institute (EBI) are gearing up to make their Sequence Reads Archive (SRA) the repository of all raw genome data24. These sequence reads represent the evidence trail for all sequence assemblies, including rare sequences or alleles that might otherwise go undetected. For smaller groups considering warehousing sample data, long term storage may not be an option, as digital storage costs will soon outweigh sequencing costs. Before long, assuming that data analysis tools can keep pace with the rapid changes in sequencing technologies, it will be less expensive to re-sequence a sample than to archive the data for a long period.

Despite all these considerations, though, the most disruptive development in this field will not be a new sequencing platform, storage solution or analytical software package – it will be the deciphering of our genetic code, which will allow us to scale down our efforts and to focus, finally, on the information in our genes that really matters.

Acknowledgement

I am indebted to Chris Schultis and Dr. Jatinderpal Deol for critical reading of the manuscript.

References

1. Sturtevant AH, Third group of linked genes in Drosophila ampelophila, Science 1913, PMID: 17833164

2. Watson JD and Crick FH, the structure of nucleic acids; a structure for deoxyribose nucleic acid, Nature 1953, PMID: 13054692

3. Min Jou W et al,Nucleotide sequence of the gene coding for the bacteriophage MS2 coat protein. Nature 1972, PMID 4555447

4. Sanger F et al, DNA sequencing with chain-terminating inhibitors, PNAS (1977) PMID 271968.

5. Lander ES et al, Initial sequencing and analysis of the human genome, Nature 2001, PMID 11237011

6. Venter C et al, The sequence of the human genome, Science 2001, PMID 11181995

7. Roche 454 (http://www.454.com/)

8. illumina Solexa System (http://www.illumina.com/technology/ sequencing_technology.ilmn)

9. Applied Biosystems (http://www.helicosbio.com/Home/tabid/ 36/Default.aspx)

10. Helicos Biosciences Corp (http://www.helicosbio.com/Home/ tabid/36/Default.aspx)

11. Complete Genomics Inc. (http://www.completegenomics.com/)

12. Ozsolak, F. et al. , Amplification-free digital gene expression profiling from minute cell quantities, Nature Methods 2010, PMID: 20639869 13. Pacific Biosystems (http://www.pacificbiosciences.com/)

14. Life Technologies Single Molecule Sequencing (http://www.lifetech.com )

15. Ion Torrent (www.iontorrent.com)

16. IBM Single molecule sequencing (http://www- 03.ibm.com/press/us/en/ pressrelease/32037.wss)

17. NABSys (http://www.nabsys.com/)

18. The Sequence Quality Control project (MAQC-III/SEQC) (http://www.fda.gov/ScienceResearch/Bioinformati csTools/MicroarrayQualityControlProject/ default.htm )

19. Conrad D.F. et al, Origins and functional impact of copy number variation in the human genome, Nature 2010, 464, PMID: 19812545

20. Personal Genome Project (http://www.personalgenomes.org/)

21. The Microbiome Project (http://nihroadmap.nih.gov/hmp/)

22. The Cancer Genome Atlas Project (http://cancergenome.nih.gov/)

23. Encode & modEncode projects (http://www.genome.gov/10005107 & http://www.modencode.org/)

24. Sequence Reads Archives (SRA) (http://www.ncbi.nlm.nih.gov/sra and http://www.ebi.ac.uk/ena/about/ page.php?page=sra_submissions )

About the Author

Bhupinder Bhullar

Bhupinder Bhullar is a Lab Head in the Novartis Institutes for BioMedical Research (NIBR) in Basel, Switzerland. He obtained his PhD from the University of Calgary, Canada and completed postdoctoral studies at Harvard Medical School and the Whitehead Institute, MIT. Bhupinder’s interest is to develop creative methods for drug discovery to bring novel targets and small molecules into the development pipeline. Bhupinder’s group employs sequencing to study genetic variation and elucidate a drug compound’s mechanism of action.

Contact the Author

email: [email protected]