Exome: A Goldrush in Clinical Sequencing

A number of new companies have recently been created, or have refocused their primary business effort on opportunities in clinical sequencing and personalized medicine. This area has received a lot of speculative attention in the past few years, but the recent development of “Exome” sequencing technology has suddenly made it a practical area for commercial investment.

There are several challenges in that must be overcome to make DNA sequencing a clinically relevant tool:

• 1) the cost of the assay, which includes sample collection from the patient, sample preparation, and operation of the DNA sequencing machine

• 2)   bioinformatics to identify sequence variants in the patient’s DNA 

• 3)   filtering and interpretation of sequence variants for clinical relevance – i.e. identify variants that provide information that directly impacts disease treatment decisions.

Exome sequencing addresses all three of these challenges. The exome is defined as the protein coding exons of genes, which make up approximately 50 MB of the human genome – about 1.5% of the entire genome. New sample preparation reagents make it possible to capture this portion of the genome in a single step in a single tube for less than $100. The current Illumina HiSeq sequencing machine produces about 20 Gbp per lane for about $1500, which is equivalent to 400X coverage of the exome. Since current bioinformatics methods require only 50-100X coverage for optimal discovery of sequence variants, this allows 4 to 8 samples to be multiplexed into a single lane. Therefore exome sequencing can be used to scan all of a patient’s genes for under $500 in sequencing and sample preparation costs. The $1000 genome is available right now.

Since the exome is a much smaller amount of sequence than the entire genome, and it is focused on the best characterized regions, the task of identifying variants is simplified. The problem of false positives is reduced both by the smaller extent of sequence and by the deeper coverage (≥50X). The challenge of interpretation is also greatly reduced since exons are by definition protein coding. All exon sequence variants can be characterized as changing amino acids or not (or creating frameshifts &/or stop codons), and the likely impact on a protein of an amino acid change can be assessed by a number of existing algorithms. Most genes can be further characterized by existing knowledge about protein function such as metabolic and regulatory pathways, as well as databases of clinical genetic and pharmacogenetic information.

Since the technical ability to perform exome sequencing and basic discovery of sequence variants is available to anyone with a HiSeq machine (and a few skilled bioinformaticians), companies are currently trying to distinguish themselves with the clinical interpretation that they can offer. Some companies are skipping the sequencing entirely and focusing solely on the interpretation of clinical sequence data.

• Ambry Genetics

Ambry Genetics is the first laboratory to provide CLIA-approved exome services for applications in clinical diagnostics along with clinical interpretation and classification of variant data. The expert bioinformatics team makes Clinical Diagnostic Exome™ possible with a robust data analysis pipeline for Mendelian disease discovery.

• Knome Offers Whole-Genome Sequencing, Interpretation for $5K

Founder: George Church

KnomeSelect, a targeted sequencing service that covers the exome, costs $24,500 for individuals. A comparative analysis of genomes includes a short list of suspect variants, genes and networks. Custom desktop software is provided for further analysis, including KnomeFinder for candidate variant discovery and KnomePathways for finding gene-gene interactions and gene networks. The company recently opened up its services to scientists interested in sequencing exomes or genomes of small numbers of humans as part of research studies.

• Personalis

Founders: Stanford founders are Russ Altman, chair of the bioengineering department; Euan Ashley, director of the Stanford Center for Inherited Cardiovascular Disease; Atul Butte, chief of the division of systems medicine at the department of pediatrics; and Michael Snyder, chair of the genetics department and director of the Stanford Center for Genomics and Personalized Medicine. John West, the former CEO of Solexa, is the new firm's CEO.

“Its core capability will be the medical interpretation of human genomes. Personalis expects to work closely with a variety of sequencing technology and service providers — including Illumina, Complete Genomics, and others.”

• Omica is a new startup company. It has developed and published the VAAST system for annotating sequence variants. VAAST is a probabilistic search tool that identifies disease-causing variants in genome sequence data. It combines elements from existing amino acid substitution and aggregative approaches that increase accuracy and make it easy to use. The tool can score both coding and non-coding variants, and evaluate rare and common variants. The platform, to be used for clinical annotations of both whole genomes and more targeted data such as exomes or gene panels, is currently in beta testing with several undisclosed collaborators. Besides VAAST, which generates disease candidate lists, the Omica service will also include annotation tools that will provide additional information about the role of the genes. Users can submit their genome sequence, and it puts all the clinical annotations on top of it. It also has an interface that can relate variants to diseases.

• GenomeQuest is a provider of cloud-based computing solutions for analysis of Next Generation sequencing data. The GQ-DxSM product analyzes and reports comprehensive genomic information about variations and changes in genes and proteins to improve disease treatment. The workflow can be used for Whole-Genome, Whole-Exome, and selected Gene Panels including:

- Automated transfer of raw data from sequencing machines

- Alignment of the reads against reference genomes

- Variant detection and annotation

- Mapping and documentation of variants against known inherited and somatic mutations

- Integration with other clinical data systems such as Electronic Health Records and therapy protocols to create a comprehensive patient diagnostic record

Designed for academic research laboratories, diagnostics labs, IVD manufacturers, and pharmaceutical companion diagnostic groups, GQ-Dx is already being used in clinical research. In collaboration with GenomeQuest, pathologists at Beth Israel Deaconess Medical Center, a teaching hospital of Harvard Medical School, are developing “clinical grade” annotation methods and databases for cancer diagnoses. GenomeQuest has also created a GeneTests-based diagnostic panel that generates a comprehensive report on disease susceptibility, diagnosis, and treatment on more than 2,000 disorders from a single, whole-genome sequence of a patient.

• Foundation Medicine has narrowed the focus even further. They provide diagnostic exome sequencing of 300 cancer related genes on FFPE tumor samples submitted by clinical pathologists. They sequence these 300 genes to very deep coverage (500X) to allow detection of rare somatic variants in heterogeneous tumor tissue. The selected gene set is intended to include only genes with directly disease related functions that impact cancer treatment decisions. The test is intended to replace many different single-gene diagnostic tests currently on the market. 

• 23andMe has started a pilot program that offers full exome sequencing for $999. While the company’s regular personal genome service uses Illumina genotyping arrays with around 1 million SNPs (single nucleotide polymorphisms), the exome sequencing actually sequences around 50 million DNA bases with 80x coverage.

Customers will get the raw data, without any additional reports, so it will only be useful to people who actually know how to handle this raw genetic data. 23andMe plans to eventually add a limited set of tools and content that utilize exome sequence data.

23andMe is not the first company to offer whole-genome sequencing to consumers, but it is the first to do so at a sub-$1000 pricepoint. For hardcore bioscientists who know their way around raw genetic data, this is as good a deal as you can currently get.

Foundation Medicine grabs for the low-hanging fruit of NGS cancer diagnostics

I was at the CHI APPLYING NEXT-GENERATION SEQUENCING conference in Providence RI, where I heard an extremely interesting presentation from a new Genomics company called Foundation Medicine. This company plans to offer a clinical diagnostic test based on very deep sequencing of all exons from about 300 cancer related genes. They will sequence directly from pathologist's FFPE blocks using Illumina HiSeq to a depth of 500 to 1000X.

Here is a recent poster they presented at ASCO, but the information at the CHI conference was updated and more in depth.
ASCO poster

Here is why I think this is very important. First, this test will include all existing genes that are currently being tested for any type of cancer (BRCA1&2, KRAS, BRAF, HER2, EGFR, etc), but will include all exons and greater diagnostic sensitivity for mutations present in low abundance in heterogenous samples which may suffer from mixed tumor and normal tissue, multiple clones, mixed aneuploidy etc. It will likely also contain the majority of known pharmacogenomic genes. So this one test could put all the other providers of cancer related genetic tests out of business.

It is also very important that the test is highly targeted only at "actionable" genes. Foundation Med. plans to deliver a report for each patient (in 14 days) that lists all mutations observed in the diagnostic genes, as well as some key items drawn from the literature, clinical trials, and a curated knowledge base about treatments relevant to those genes. In the presentation, COO Kevin Krenitsky said that they typically found 2-3 mutated genes per patient. This is an amount of data that the oncologist or pathologist can reasonably be expected to deal with — rather than the hundreds to thousands of mutated genes with questionable to zero clinical implications that will be produced by whole genome sequencing.

Another interesting discovery reported by Foundation Med. was that in a small number of cases (perhaps 5%), they found mutations for genes that were associated with a different type of cancer. This suggests the use of a non-traditional drug, possibly in combination with other more typical therapies, as an individualized treatment for that one patient. There are currently about 30 drugs for which genetic information can aid in treatment decisions, but this is clearly an area of intense development. Foundation Med. can easily modify its test to include any relevant new genes. We are clearly heading to the point where every cancer patient will benefit from an individualized genomics workup.

GWAS vs Exome Sequencing

I learned something interesting today about the SNP arrays used for GWAS. There has been a lot of discussion about the nature of mutations/alleles discovered by GWAS studies in terms of the "common disease: common variant" hypothesis. It is clear that SNP arrays are designed to cover common variants - alleles that are present in at least 2% of the human population (or at least of some population). Contrary-wise, genome sequencing studies tend to focus on rare variants. In fact a number of recent studies show that major diseases such as cancer and autism tend to be associated with novel, very severe mutations in coding regions of genes.

Now this is the interesting part. We took a look at the intersection between the Illumina 2.5 M SNP array and the regions targeted by the Agilent Sure Select exon enrichment kit. It turns out that only about 90K of the Illumina SNPs are in the exon regions. This matches up with Illumina's own annotation file showing that more than 80% of the SNPs on the array are intron or intergenic.  My human genetics colleague suggests that the SNP array targets sequence variants (alleles) with small effects, while the exon sequencing strategy targets mutations with large effects. So we can't really replace the SNP array with exome sequencing, they are looking at completely different things.

The False Discovery of Mutations by Sequencing

     I am amazed by the success reported in recent papers finding mutations by Next-Gen Sequencing in rare genetic diseases and cancer. In our lab, the sequence data for SNPs is messy and difficult to interpret. The basic problem is that NGS data, particularly Illumina data in our case, contains a moderate level of sequencing errors. We get somewhere between 0.5% and 1% errors in our sequence reads from the GAII and HiSeq machines. This is not bad for many practical purposes (ChIPseq and RNAseq experiments have no trouble with this data) and this error level "is within specified operating parameters" according to Illumina Tech support. The errors are not random, they occur much more frequently at the ends of long (100 bp) reads. Some types of errors are systematic in all Illumina sequencing (A>T miscalls are most common), and other types of errors are common to a particular sample, run, or lane of sequence data. Also, when you are screening billions of bases, looking for mutations, rare overlaps of errors will occur.
     So if sequence data contains errors, and the point of your experiment is to find mutations, then when you find a difference between your data and the reference genome (a variant), you had better make doubly sure that the difference is real. There is a lot of software designed to filter out real mutations (SNPs) from the random sequence errors. The basic idea is to first filter out bad, low quality bases using the built-in quality scores produced by the sequencer. Second, require that multiple reads show the same variant, and that the fraction of reads showing the variant makes sense in your experiment: 40-60% might be good for a heterozygous allele in a human germline sample, 10% or less might make sense if you are screening for a rare variant in a sample from a mixed population of cells.  Also, it is usually wise to filter out all common SNPs in the dbSNP database - we assume that these are not cancer causing, and they have a high likelihood of being present in healthy germline cells as well as tumor cells.
     We have used the SNP calling tools in the Illumina CASAVA software, the MAQ software package, similar tools in SAMtools, and recently the GATK toolkit. In all cases, it is possible to tweak parameters to get a stringent set of predicted mutations, filtering out low quality bases, low frequency mutations, and SNPs that are near other types of genomic problems such as insertion/deletion sites, repetitive sequence, etc. Using their own tools Illumina has published data showing a false positive detection rate of 2.89% (Illumina FP Rate).  Under many experimental designs, validating 97% of your predicted mutations would be excellent.
     Unfortunately, our medical scientists don't want predicted SNPs vs. an arbitrary reference genome. They want to find mutations in cancer cells vs. the normal cells (germline or wild type) of the same patient. This is where all the tools seem to fall apart. When we run the same SNP detection tools on two NGS samples, and then look for the mutations that are unique to the tumor vs the wild type (WT), we get a list of garbage, thousands of lines long. We get stupid positions with 21% variant allele detected in tumor and 19% variant in WT. Or we get positions where the 80% variant allele frequency is not called as a SNP in WT because 2 out of 80 reads have a one base deletion near that base. So the stringent settings on our SNP discovery software create FALSE NEGATIVES where we miss real SNPs in the WT genome, which then show up as tumor-specific mutations in our SNP discovery pipeline.
     Zuojian Tang is creating a post-SNP data filter that imposes a sanity check on the data based on allele frequencies. We are trying out various parameters, but something like a minimum of 40% variant in the tumor and less than 5% variant in the WT narrows the list of tumor-specific mutations down to a manageable number that could be validated by PCR or Sequenom.

Involve Bioinformatics in design of every experiment... please

Two interesting projects came through our informatics group last week, both in the 'data drop' mode were the investigator asks for help to analyze data as it comes off of the sequencers. I have noted many times before, that our informatics effort is much greater on the poorly designed and failed experiments.

Experiment #1 was a seemingly standard SNP detection using exome sequences with 100 bp paired-end  reads on Illumina HiSeq (Agilent Sure Select capture) - the entire thing done by an private sequencing contractor. The contractor also supplied SNP calls using Illumina CASAVA software. Our job was simply to find overlaps between the SNP calls for various samples and controls, and to annotate the SNPs with genomic information (coding or non-coding, conservative mutations, biological pathways, etc).  However, we have an obsession with QC data, which the vendor was very reluctant to supply. Turns out that these sequencing reads have a 1.5% error rate, while our internal sequencing lab generates 0.5% error. We also see 10K novel SNPs in each sample with only minimal overlap across samples (a red flag for me). More QC data is extracted from the vendor, and now we see a steep increase in error at the ends of reads. So we wish to trim all reads down by 10-25% and recall SNPs - extract more files from vendor 3x (Illumina requires a LOT of runtime and intermediate files in order to run CASAVA for SNP calling).

Meanwhile, Experiment #2 is an RNAseq project where the investigator is interested in alternative splicing. We analyzed one earlier data set with 50bp reads with only moderate success. It seems that very deep coverage is needed to get valid data for alt-splicing, especially when levels of a poorly expressed isoform are suspected to change by a small amount due to biological treatment. The investigator saw some published results suggesting that paired-end RNAseq data would provide more information about splicing isoforms. So, WITHOUT a bioinformatics consult, they sent an existing sample (created for 50bp single end sequencing) to the lab for 100 bp paired-end sequencing. This data came out of our pipeline with more than 20% error and a strange mix of incorrectly oriented read pairs (facing outward rather than inward). After a few days of head scratching and escalating levels of Illumina bioinformatics tech support, we have an explanation. A 225 bp library fragment contains 130 bp of primers and adapters. Thus the insert has an average size of about 95 bp. Some are shorter!  Thus, our 100 cyle reads go off the far end of most sequences, adding 5 or more bases of adapter sequence where the alignment software is expecting genomic sequence. In addition, the paired ends overlap more than 100% - so the start of one read is inside the end of the other. Thus they map in the opposite orientation, with an insert size of 5-10 bp. Our best effort to analyze this data will involve chopping all reads back to 36 bp and repeating the Paired-End analysis. So that was 3 days of bioinformatics analysis time not so well spent on forensic QC.

Now we are looking back to Experiment #1 and wondering about insert sizes in that library. What if that library's insert size was about 110 or 120 bp (perhaps with a sizeable tail of much smaller fragments), and a fraction of the reads also run off into the adapter, adding mismatched bases at the ends of alignments, and thus jacking up the overall error rate.

Two conclusions: 1) talk to bioinformatics BEFORE you build your sequencing libraries
2) if you want something done right, do it yourself.