Advances in Genome Biology and Technology Conference 2010

I'm just back from Florida and decompressing from the Advances in Genome Biology and Technology (AGBT) Conference. It takes place every winter (this is the 11th year) on Marco Island, which is somewhat inconveniently located about an hour's drive from the Fort Myers airport. I flew direct from Baltimore but a lot of people had to connect from Charlotte, Chicago and other places.

This is the first time I've attended this meeting, but it reminds me of the Genome Sequencing and Analysis Conferences (GSAC) from 10 or 15 years ago. Those conferences were in Hilton Head, South Carolina, among other places, which was also far from an airport. When the Human Genome Project was getting underway, GSAC was an exciting blend of science and technology with a large vendor presence.

AGBT doesn't have vendor booths like GSAC used to, but there were hospitality suites and a good number of vendor-sponsored 'workshops' which could cynically be called advertisements, but which in most cases were pretty informative.

There's no way for me to cover all the developments in one blog post, but I'll give a few highlights. Illumina maintained its market lead by rolling out a new sequencer, the HiSeq 2000. It's the same basic technology as its earlier Genome Analyzer line, but they can now do good coverage of two human genomes in a single run for $10,000. Roche/454 seems to be getting marginalized and didn't have much of a presence. Helicos was still present with their single-molecule sequencer but appears to marred by high error rates. Life Technologies is also improving its SOLiD platform to increase throughput and is nipping at Illumina's heels. There was no mention at all of the Polonator.

But the new technologies were the ones that dominated the meeting. Life is working on enhancements to the 2-base encoding on the SOLiD platform that they claim will improve accuracy and also stop reporting data in colorspace, which is often confusing. Complete Genomics has delivered on their promise from last year's AGBT to start delivering full human genome sequences, of apparently good quality, from their sequencing factory (they don't sell sequencers, they sell the sequence data).

Pacific Biosciences had a workshop where they talked about their new sequencer. It uses DNA polymerase molecules to incorporate labelled nucleotides on a single DNA molecule, detecting each incorporation to reveal the sequence. They also discussed some neat variations, such as resequencing the same molecule multiple times to get higher quality data, and producing so-called strobed reads, which allow reading sections of longer molecules, giving the same benefit as paired-end reads for spanning repeats and identifying structural variations. Their machine also has the benefits of being able to generate longer reads (up to 10,000bp, but only on some reads in a run) and being able to do runs in hours instead of days. But unlike Complete Genomics and the more established companies, they didn't have any reports from collaborators that used their system, and there were questions about the accuracy of their sequence which didn't really get answered.

But the real fireworks happened in the last two talks of the conference. The first one was from Life Technologies, reporting on a new kind of sequencer they hope to roll out within the next year. This one uses DNA polymerase tethered to a nanoball that emits light in a specific frequency that excites dye-labelled nucleotides as they are incorporated. They had several cool movies to demonstrate the idea. Like Pacific Biosciences, this method promises fast run times and potentially long read lengths. However, in this case the polymerase is free, not in a tiny well, so that once it stops working, it can be replaced with another one. This means the read lengths might not be as unpredictable. It means they can also resequence the same molecule multiple times for higher accuracy. Also, the machine itself is quite small. This is clearly interesting technology but early in the commercialization process.

Lastly, Ion Torrent described their new sequencer, which should be available in the first half of this year. In some ways it is similar the earlier 454 machine, and in fact Ion Torrent's CEO also worked on the 454 machine. Ion Torrent's device is also based on using DNA polymerase but they made a huge deal out of the fact that their device is based on silicon semiconductor technology. They capture the positive charge generated by incorporation of one (or more) nucleotides and convert that to a digital signal on their chip-like device. Their device is small and promises fast run times, but they didn't talk about sequence quality or the sample preparation process. The biggest distinction they made was that they could make the devices very cheap by relying on existing semiconductor fabrication technology and existing factories.

This is by necessity a brief overview, and I hope to post more entries as the community processes all the information from the ABGT meeting and some of these new machines become more widely available. But it was an exciting, inspiring, and somewhat overwhelming meeting. I think the biggest challenges will be to figure out which machines are best for which applications and find good ways to process and understand the huge volumes of data that will be generated.

Many feet, many paths to EHR

When Bill Clinton had stents placed in one of his arteries a few weeks ago, our colleague Christine Quern (Feinstein Kean Healthcare) mentioned that she’d heard a physician commentator on the local newscast state that Clinton went to Columbia Presbyterian Hospital for treatment “because that’s where his medical records were.”

I’ll forgive a news commentator for inaccuracy under fire. But in this era of “health IT,” with intense focus and funding for creating electronic medical records, is this kind of thinking old fashioned? Is it acceptable to state that someone should seek treatment where their medical records reside? Even a former president?

Well… yes. Many strident cases are made for “digitizing” medical records through the wounded warrior stories, in which families of injured veterans cart around suitcases of medical records from doctor to doctor, seeking treatment and answers. This is one of the business cases that drives the Nationwide Health Information Network (NHIN), which seeks to provide the trust fabric for secure electronic health information exchange. One of the promises of the NHIN is that you can leave your folder/binder/suitcase at home, because Dr. Smith can obtain the test results taken by Dr. Joshi. It's a fantastic vision of a primary benefit of electronic health records.

I work with the wonderful people behind the NHIN; I work to create the testing and validation processes that will ensure that the systems exchanging health information across this network are capable of doing so. I’ll write more about the work being done, but I work on the NHIN and had to be my own personal NHIN at my doctor’s office last week – Dr. S---  faxed my test results to me, and the e-fax rendered a pdf on my laptop, which I showed to Dr. P--- in his office. I work on the NHIN and realize we’ve got a long way to go before its promise can be realized – we’ve got a long way to go before it would be considered absurd that Bill Clinton would get treatment at a hospital just because his medical records were there.

Why aren’t we there yet? Government, and businesses big and small, are attempting to clear the path to make the sharing of health information simpler – everyone from Dell to seemingly Pizza Hut is getting into the effort. We’d like to hear your thoughts on the largest challenges – time, money, technical hurdles, privacy concerns, workflow modifications, access consent, human nature? – how will conditions unfold so that Bill Clinton can access his medical records anywhere? What will that look like – will Clinton own his information (in a chip in his arm, in his iPhone, in HealthVault or Google Health et al, or in his suitcase full of medical records), or will the data reside where it originated, shareable with others? Is our first step defining standards for machines to exchange data, or defining up a way for people to share information? We’re heading to the Health Information Management Systems Society conference (HIMSS) next week and will report on our conversations, musings, and discoveries. In the meantime, please share your own comments and ideas.

What Gets You Up at 5AM? Focus: Rare Disease Community

As part of 5AM's deeply ingrained commitment to understanding the facets of the community we seek to serve, this marks the first of a series of entries focused on learning what gets our clients, partners and the collective set of biomedical stakeholders up at 5am!

The team at 5AM was thrilled to participate in the series of meetings organized by Genetic Alliance and the NIH's Office of Rare Disease Research. The volume and diversity of people so willing to share their perspectives, ideas, operational models and concerns surrounding rare disease issues and research was simultaneously informative and inspirational. The sense of urgency, combined with considered approaches, gives birth to a vision that can drive change now and for generations to come. We appreciate that the use of registries, the exposure of biorepositories and the protocols that inform them, and the ability to associate consented data - be it patient reported such as Family Health History - or provider collected clinical, imaging or molecular data - will lead to a greater understanding of disease. We are optimistic this will ultimately provide the means to inform treatment and hopefully, cures.

After demoing our Biospecimen Locator at the Registry and Repository Boot Camp, we sat down to figure out how best to collaborate with the variety of forces - government, non-profit, commercial, and patients and their families - to move the collective, burgeoning vision forward, focusing on what can we do today. The Biospecimen Locator enables the exchange of information to facilitate research by moving specimens from the freezer to the hands of researchers.

We hope you will consider leveraging the collective work we've done with a wide variety of stakeholders as we all look to harmonize standardized common data elements, vocabulary, and open source software to facilitate research collaborations that produce results. Please take advantage of our offer to do a demo of this open source software that can be used today to simplify the visibility and exchange of biospecimens.

We would like to hear from you - please continue to share your thoughts and needs. We are consciously collaborative, and in order to support moving a collective vision forward, we need your voice and input and will offer our own. We'll be adding resource links here and will closely follow the trajectory of this space.

I'll close with one of the brilliant quotes in a breakout session from Dr. Carolyn Compton, Director of the Office of Biorepositories and Biospecimen Research, when talking about how to propel change:

1. You can make people do it
2. You can pay people to do it
3. You can show people VALUE and they will do it themselves

Building upon the standing ovation she received - Bravo! 5AM is proud to join this community and expect to contribute technology solutions to the Value Chain discussed.

Extreme Visualization Makeover #1: Genome.gov’s “Published Genome-Wide Associations” Chart

Welcome to the first installment of Extreme Visualization Makeovers! In each installment, we’ll look at a different data visualization, chart, or graphic from the literature or the Web, and see if we can find ways to make it more effective. While each installment may not prove to be “extreme” – that makes a better title for our series, so we’re sticking with it.

Our first subject provides us with a really great teaching moment on color semiology. Semiology is a favorite term-of-art in visualization – simply put, it means “pertaining to communication through signs and symbols.” We use it to mean any system of signification – whether it’s through icons, color coding, numbering systems, mapping symbols, etc. Semiology encompasses the art and science of choosing the right way to signify things.

Which brings us back to the first subject of our series. Genome.gov publishes a quarterly summary graphic showing the loci of all the SNP-trait associations with p-values < 1.0 x 10-5, plotted as colored dots on a graphic representation of the human chromosome complement.

click to view full-size PDF from genome.gov

Now, this chart has grown over time to encompass 104 such traits (as of this writing). The authors of the chart have chosen to differentiate the traits through color semiology – each trait is assigned a unique colored dot. Take a moment to view the full-size graphic (click on the thumbnail above – the chart will open in a new window). Now – see if you can uniquely identify all of the orange dots, and the traits they represent. It’s pretty tricky, isn’t it? I actually had to resort to Photoshop’s color-sampling tool to tell some of them apart. It’s virtually impossible for a fully-sighted individual – imagine how tough this chart is for someone with color-compromised vision.

The problem here is that color semiology is not appropriate for such a large value range. We simply don’t do well differentiating 104 different colors from a field of dots. Compounding the issue is the effect of the Gestalt Color Principle – our brains want to group together things with really similar colors, which can be useful in some instances, but here it just makes matters worse.

In 1969, Brent Berlin and Paul Kay published a groundbreaking study of color perception across culture, in which they proposed that there were really 11 fundamental (or “focus”) colors that everyone could easily differentiate, most likely based on some underlying physiological or neurological principle. The Berlin-Kay palette was extended to include cyan by the visualization guru Colin Ware, giving us 12 colors that are reasonably safe to use for ordinal color semiology in infographics and data visualization. What do I mean by “ordinal” color semiology? Data dimensions that are sets of things (like categories, without quantitative interrelationships), rather than continuous, ordered, quantitative values, are ordinal. We can also use color for quantitative values – in fact, we can even split color into its three component subdimensions – hue, saturation and value – and use each of these to represent a separate quantitative dimension. Heatmaps and terrain relief are examples of such quantitative color semiology. We still have to be careful though, because we’re fairly bad at discerning specific quantitative values in a color (hue, saturation, or value) range.

But in the graphic we’re considering here, the authors are trying to use ordinal color semiology for 104 separate ordinal values. By now, you should understand why this is disastrous. It’s nearly 10 times as many colors as the Berlin-Kay set. It’s a set-up for failure.

So, how might we improve matters? One approach might be to employ a hybrid semiology – for instance, grouping the traits into manageable sets (with 12 or fewer sets in total) and encoding these sets with color semiology. Then, within each set, numbering the traits (numeric semiology). Let’s see how this might work, using chromosome 18 as a guinea pig. First, here’s what we’re starting with (excerpted from the original document):

Notice how your eyes and mind have to work to make sense of this, even though it’s just one chromosome from the whole diagram – you can do it, but it’s not intuitive or fast. Also, notice that the two Type 1 diabetes dots may actually look slightly different in color, due to their proximity to dots of different colors – this kind of color interaction is another hazard of using lots of different colors jumbled together to represent things. If there were only 12 well-differentiated colors on the diagram this would not be as big of a problem, but on the full diagram with 104 colors, there are too many things that are “lavender-mauve-ish” – so these kinds of visual effects become meaningful.

Now, let’s rework it a bit by sorting our traits into categories, and assigning a “Berlin-Kay safe” color to all traits in the same category. Then we’ll number within each category and put the numbers on the dots.


Suddenly, you can find things! It works well in both directions – whether you start from the legend or from the loci on the chromosome. This solution will scale up to quite a large number of traits without losing its efficacy, as long as the number of categories stays at 12 or less.

Is this the only solution (or even the best) to this problem? Probably not. There are other possibilities as well – one approach would be to split the diagram into multiples – duplicate copies of the whole diagram, broken out by some value, such as the categories assigned above (i.e., a diagram showing only cancers, another showing only cardiovascular loci, etc.). A possible disadvantage to this approach would be that you would no longer see the proximity of seemingly unrelated traits on the same chromosome, which might hinder insight into linkages, etc.

I hope this stimulates your thinking about choosing the right way to signify things. Please comment – do you see another way to approach this? Do you think I’m way off base (or right on)? Also, if you encounter any other charts, visualizations or infographics that you think could use a makeover, please send me a link and I’ll add them to the list for consideration.

Thanks for reading, and be sure to check back here for future installments of Extreme Visualization Makeovers!

Done, Done, Done, and Done

In the agile world, the definition of done tends to center around backlog items and the iteration boundary. The difficulty I see with such definitions is that they are not granular enough for individual developers. Development work proceeds commit by commit, not user story by user story. Our best practice is for a developer to commit at least once a day, so we don't lose work to vagaries such as hard drive failures or accidental deletion. This has led us to multiple definitions that break down work from the user story to each commit. At 5AM, code can be "Done" for 4 purposes, of increasing rigor. (1) Commit to a private branch, (2) commit to mainline development branch, (3) resolving a subtask, (4) resolving a backlog item. Here's what each definition entails:

Private branch commits: Commits to private branches can be made at any time, without restriction. The code doesn't even have to compile. This permits daily checkpointing and keeps us from losing work.

Mainline development commits: Commits here affect other team members, so we have several requirements. Every commit must be tied to a tracker item. Peer review is required, no matter how minor the change. Code must pass continuous integration (CI), which brings in things such as coding conventions and the regression suite (and the sombrero). These requirements ensure traceability, that CI always passes, and that at least two eyes have seen every piece of code.

Resolving a subtask: Individual tracker items are the lowest level of granularity we typically talk about at daily standups. Resolving subtasks requires not only committed code, but high quality unit tests, removal of fixmes in the code, and known limitations or escapes are captured as additional subtasks. We don't measure velocity here, but we have found that the team gets a good sense of progress by watching subtasks move to resolved.

Resolving backlog items: This is where most scrum definitions of done center, and where the team measures velocity. Our definition is similar to others - beyond resolving all subtasks, resolving a backlog item requires that: Functional and non-functional requirements have been met. Integration tests exist. Escaped or deferred functionality is captured. In totality, the code is deployment-ready.

We're done, done, done, and done.

Mammography Screening - Don't Panic!

Surely you saw the newspaper articles about the recent government guidelines about breast cancer screening using mammography. There were breathless articles and letters to editors complaining about how mammograms had saved lives and how could a heartless government agency say we shouldn't continue with the practice. Congressional hearings were held with "debate split along party lines" so you know they really got down to the science of it.

The main idea that caused so much consternation was that women aged 40-49 should not routinely undergo mammograms to look for signs of breast cancer, which is what previous guidelines had said. But let's look at what these new guidelines really say:

• "The USPSTF recommends biennial screening mammography for women aged 50 to 74 years"
• "The decision to start regular, biennial screening mammography before the age of 50 years should be an individual one and take patient context into account, including the patient's values regarding specific benefits and harms"

So it doesn't say that women in their 40's should not get mammograms. It says they should weigh the risks and benefits themselves (and with their doctor, obviously, although I wish it had said that explicitly) and make their own decision.

So why not get mammograms earlier? The term 'screening' is key here. A screening test is a test that is given when there's no prior evidence or risk for a condition. A cholesterol test is a screening test, for instance, to look for evidence of heart problems.

You have to remember that by its definition a screening test is given to large number of people, only a small fraction of which have the condition being tested for. So even if such a test is relatively accurate, there will be a large number of false positive results. In the case of mammography, a false positive (and I hesitate to use the word 'positive' here) is an abnormal result when in fact the patient does not have cancer. There's a useful statistic to quantify this called positive predictive value (PPV). PPV is the fraction of positive results that actually have the condition being tested for. For women aged 40-49, mammography has a PPV of 2-4%. So that means that only a small percentage of women in that age group who have an abnormal mammogram actually have cancer.

The critics of these guidelines have said this is fine since we want to catch as many cancers as early as possible. But you have to take into account that no test is without risk, and that more invasive procedures, such as biopsies, are done when a mammogram is abnormal. The mammogram itself, as well as the subsequent procedures, have a monetary cost as well.

What the guidelines say is that women under 50 should get mammograms if they want to, or if they or their doctor feels there are other reasons for them to have a high risk of breast cancer. If one has a family history of breast cancer, or a genetic risk factor, then those would be good reasons to have earlier mammograms, in my non-medical opinion.

So this is a case of reasonable scientific conclusions being misinterpreted. However, I do think there needs to be more research done to say how many cases of cancer would be missed if mammograms were only done for women aged 40-49 who had some other risk factor. That would make it more concrete for women about what risk they were running by not having mammograms before age 50.

The Case for Complexity

I enjoyed reading Neil Saunder’s blog post "Has our quest for completeness made things too complicated?" last week. In the blog, Neil advocates that capturing –omic data in a simple feature-probe-value triplet is a much more practical approach to the data integration problem than trying to work with data across a multitude of ontologies and schemas that have been developed for individual data types. Neil’s focus is to identify what’s the same across all these data sets and work with that very simple model rather than spend a lot of effort to come up with a complete model of everything.

So, to you Neil Saunders, mostly I say “Amen, brother! I’m with you.” My physics background also gravitates me towards boiling a problem down to its essence – the bare facts upon which complex analyses can be built. As much as I believe this, however, I also believe that ontologies and other meta-data play an essential role in integrative analyses. Let me illustrate…

We recently were working with a client to integrate multiple –omic data sets (gene expression, GWAS, proteomics, etc). Similar to Neil, our first step was to extract the data from their individual repositories, ontologies, schemas, etc and boil these very disparate data sets down to their greatest common divisor (GCD). “Greatest common divisor” is an apt term here because in selecting what data was included in our simplified model we looked both for what was “common” to all data types and the analyses that would follow and also what was “greatest” in the sense that wherever possible we wanted to take processed results over raw data. In Neil’s case, his GCD was the feature-probe-value triplet. For us, the GCD was a combination of probe, sample and value. It would take another blog post (or more) to go into this project and design decisions, but for now I wanted to emphasize the similarity with Neil’s approach and perhaps holler another “Amen!”

But hold on! Before tossing all ontologies onto the scrap pile of failed good ideas (somewhere between the Apple Lisa and Pepsi Clear, I imagine), let’s take a step back. After extracting the multiple, disparate data out of their various schemas and into a single, clean model, what can we do with it? The answer: not much. In our project, although the data was all in the same form, the domains were still very different. How do we relate the value we got from a SNP to one from a transcript? Should data for a sample labeled “breast cancer” be grouped with data from samples labeled “human mammary carcinoma”? The truth is although we had a unified data model, the data was far from being integrated.

So what did finally bring us to an integrated data set? Ontologies, thesauruses, mappings, etc. In our case we used multiple sets of mapping data to take our primary data values from probes to a common feature – a gene (Entrez). We used nomenclatures like SNOMED and MeSH to normalize our samples. Only after leveraging this ontological information could we work with the data in any meaningful way.

It didn’t stop there. Once we had the data mapped to common feature and sample terminology, we then utilized ontologies like GO to form and test hypotheses on biological function. Our plan going forward is to leverage disease ontologies, pathway information and other meta-data to go beyond simple lists of differential features and go towards true biological understanding. Although coming up with a good ontological model capturing the entities and relationships relevant in biology is hard and complex, it is still a good way to capture our biological knowledge in a form that can be directly applied in analyses.

Finally, lest we forget, although dealing with a different schema/ontology for each type of data is annoying, it is far better than the alternative of having no such schema/ontology in place. This was painfully clear in our project when dealing with the relative uniformity of gene expression data in repositories such as GEO compared to the state of proteomic and even GWAS data.

To sum up, I agree Neil – you’re on to something. Getting your different data sets into a simple, consistent model is the way to go. We shouldn’t try to build a complex schema/ontology to record all things for all data. However, once you’ve got the data in this simple form and are ready to move on to the analysis, I think you’ll find the ontologies to be indispensible.