At CosmosID, we are extremely fortunate to work with the founder of our company, Dr. Rita R. Colwell. Just a few of her credentials include: 11th Director of the National Science Foundation (first female director), recipient of the 2006 National Medal of Science of the United States and the 2010 Stockholm Water Prize. She is a member of the National Academy of Sciences and has authored or co-authored 17 books and more than 750 scientific publications. Her research is focused on global infectious diseases, water, and health, and has affected people across the globe, especially in regions where cholera and other waterborne diseases are endemic.
We decided to ask Dr. Colwell some questions, so we could learn more about her and how she became such an eminent environmental microbiologist and scientific leader. We learned that she wrote the first computer program to do microbial systematics with marine bacteria – on the IBM 650.
Here is an example of what Dr. Colwell created, one of the first heat maps in biology representing the phenotype similarities between 93 strains of flexibacteria:
Now, fast-forward to 2015, and we are creating similar figures at CosmosID using genomic information rather than phenotypic data only … and Mac laptops rather than an IBM 650.
Poorani Subramanian (PS) and Kelly Moffat (KM) interviewed Rita Colwell (RC).
Interview with Dr. Rita Colwell Part I
Tell us about what motivated you to establish CosmosID. You have said before that it had to do with the 2001 anthrax attacks. I also remember you mentioning a family member who had become very ill with an infectious agent and it took a long time for the problem to be diagnosed. Was this a motivator for moving into the clinical space?
One of the motivations for establishing the company was the anthrax event, essentially the final straw in a series of events. Things really had to change. There was a need for rapid diagnosis because of deaths in people who were sickened by anthrax. Diagnosis took weeks. Because I'd done so much work in systematics it was obvious to me that a rapid diagnostic was urgently needed. The clinical importance became so dramatic and also personal because a close relative died, essentially from a misdiagnosis. Thus, it was an inspiration to move very quickly and strongly to develop rapid and accurate clinical diagnostics.
For my PhD thesis I had written the first computer program for microbial systematics of marine bacteria, that is, to identify bacteria to species using phenotype, which was slow and tedious. That was a time when we grew the bacteria in the laboratory first and then tested each isolate for a variety of traits, such as ability to ferment carbohydrates, to utilize various amino acids, and react to other tests, including staining and production of specific metabolites. The results were coded and the computer program was employed, which used a similarity index.
When the genomic revolution went into full swing it became clear to me that microbial systematics should include modern molecular biology because we needed rapid, accurate identification of microorganisms. The final trigger to really do something in the clinical space was when my brother in law developed a serious leg infection that became complicated to diagnose. At first the diagnosis was Mycobacterium fortuitum, but subsequently the diagnosis was something else. He ultimately had to have his leg amputated and when he succumbed to the infection, it was determined that the causative agent of the infection was MRSA. It was very unfortunate and tragic and unnecessary, because had the diagnosis been accurate and rapid from the beginning, treatment could have been aggressive and effective.
Now that tragic outcome has been echoed so many times in talking with many people, including colleagues within CosmosID and friends for whom the clinical diagnosis was inaccurate and took a long time to make. In the best of circumstances, traditional microbiology requires time to grow the bacteria, usually 24-48 hours and then more time to determine antibiotic resistance, which can require another 24-48 hours. What usually happens is the physician makes the initial diagnosis based on symptoms, essentially his or her experience with the disease and prescribes a generally useful antibiotic. However, the actual identification and determination of resistance or sensitivity may take a week or more. By that time either the patient has recovered or a secondary infection has occurred. Very often when treating viral infections, which are not susceptible to antibiotics the bacterial flora is disrupted by the antibiotic and a secondary infection will often occur. In that case, the patient is treated with an antibiotic that shouldn't have been provided at all and the patient can end up with a double dose of antibiotics that disrupts the normal microbial flora.
Could you start at the beginning? How did you start your computer work? When you were in graduate school you were studying marine microbiology. How did you start using computer science to do your work?
The trajectory was complicated in that my plan after I graduated from Purdue University with a degree in bacteriology was to go to medical school and I was accepted at Yale, Western Reserve, and Boston University (the medical schools I had applied to), and was accepted at all three. Thus, I was on my way to medical school when I met my husband, Jack, and we got married just before my graduation from Purdue. He was a graduate student and had returned from service in the military and had started his first year of graduate school. We didn't want him to lose that year so we planned to stay at Purdue for him to get his master’s degree and I would study for an M.S. degree and then go to medical school.
As it turned out Jack was accepted at the University of Washington to do graduate study in physics and I applied to the medical school. But the University of Washington, even to this day, has a rule that medical students must be legal residents of Washington State, Oregon, Idaho, or British Columbia. So I ended up in graduate school in microbiology. When a new professor arrived about the same time from Scotland, he started a marine microbiology laboratory and I became his first student.
John Liston, the professor, was working on marine bacteria and that was an interesting new field. Since I couldn't go to medical school because I was not a resident of the state of Washington, I pursued my PhD and subsequently became well recognized for my work on marine bacteria, namely Vibrio species.
I was the first to isolate Vibrio parahaemolyticus from the natural environment in the United States. Vibrio parahaemolyticus is a very serious, seafood-borne pathogen first discovered in Japan. When I finished my graduate work, I post-doc’d with Dr. Norman Gibbons in Ottawa, Canada. Jack and I had both been awarded National Research Council fellowships. But the time being what it was, I received a second letter that said because of nepotism rules, a husband and wife could not both have a fellowship, only one or the other.
Dr. Gibbons very kindly said, "Well, come to the National Research Council here in Canada anyway" and he made available anything I needed from the stockroom, and provided a laboratory. Dr. Liston suggested we apply to the National Science Foundation so I could have my own grant and pay my own salary, which Dr. Liston and I did. I was named the PI and he arranged for me to be appointed a research assistant professor at the University of Washington. So I showed up in Ottawa with my own grant, with a lab and supplies, and it worked out well. After that I was recruited to a faculty position at Georgetown University in Washington, DC.
At a local American Society for Microbiology meeting in Washington, DC, I presented my work on the occurrence of Vibrio parahaemolyticus in the Chesapeake Bay. Dr. John Feeley, a scientist at the NIH, was in the audience and after the lecture asked, “Why aren't you working on Vibrio cholerae?” And I replied, “Well, I don't have any isolates to study”, and he said, “I’ll send you some.” He sent me a dozen strains and the rest is history. Henceforward, I was an expert in Vibrio species, including Vibrio cholerae and subsequently was asked to do research in Bangladesh and have worked in Bangladesh since 1975.
When you were doing your graduate studies and even beyond - you were doing a lot of work on the computer. Was that very common for your field of research then? Because I was looking for papers from the era when you first published and there were very, very few computational studies in marine biology or microbiology. How was everybody else doing systematics of microorganisms?
What had happened was that taxonomists studying plants and animals had begun using computers to handle phenotypic data. I was also trying to identify bacteria in the ocean and associated with marine animals. Peter Sneath in England suggested a similarity index could be used to match up bacteria to determine in a more quantitative way how strains were related, based on Robert Sokal’s work in Kansas with Dr. Charles Michener who had developed an approach to quantitate relationships amongst bees, other insects, plants, and animals. Peter Sneath recommended that something similar be done with bacteria. I had read his papers, as had my professor, and we decided that was the way to begin the modernization of microbial systematics and taxonomy.
The University of Washington had just installed the first of what was called a high-speed computer, but by today's standards it was probably the amount of computation you'd have in your wristwatch! It was the IBM 650. It was installed in the attic of the chemistry building at the University of Washington and students were allowed to use it between midnight and 6:00 a.m. There were no technicians, no operators. You operated it yourself and you had to wire a board for your own application every time you used it. I worked between midnight and 6:00 am each day. I wrote a program in numerical code, which was an interesting challenge. Furthermore, the actual computation required an hour or two to run similarity indices for 40 strains of bacteria and 78 characteristics based on phenotypic traits such as amino acid and carbohydrate utilization. I was a pioneer in writing the first program to identify bacteria straight from the marine environment, i.e. seawater, fish, algae and invertebrate animals, and all of this current work began from that simple approach.
It’s really foreshadowing if you look at CosmosID now, it's an interesting tie-in.
It’s directly correlated. What has happened is that I’ve been able to convert from phenotype to decoding the genome itself. And that to me has been really exciting. And what’s been even more fascinating is that it was expected that with greater precision, it would completely wreak havoc with bacterial taxonomy, but it hasn't. It has emphasized the general accuracy of bacterial taxonomy and clarified relationships, so the accuracy of identification has improved, but it hasn't really shifted microbial taxonomy drastically.
Were the programs you wrote used for a long time?
Oh yes, the programs were shared with many investigators who used "the little bug program" I had written, which migrated like most insects do, around the country, and I did a lot of analyses for those who didn't have access to computers.
Where you started, which was so far ahead of its time, has come full circle.
What’s interesting is that I worked on a computer that was physically a huge instrument, and sluggish and very limited in its capacity by today’s standards. Now we work on a laptop or a server, with far greater capacity, far greater accuracy. But the fact that this can be done in the laboratory without having to trot all the way across campus to a computer center and reconfigure a control board and to be able to transmit information electronically through the Internet is quite amazing.
You were ahead of your time with your first computer program and it's similar with metagenomics. You started CosmosID in 2007, when whole genome shotgun sequencing used for metagenomics was in its infancy. How did you predict that metagenomics was going to go in this direction?
Again I was one of the pioneer microbial ecologists. It’s very interesting that at that time, 40 or 45 years ago, taxonomy was considered dull, boring, and a necessary nuisance. Similarly with microbial ecology, it was considered a very imprecise field, with unclear ideas of the relationship of microbial distribution patterns with environmental drivers. And having worked on marine bacteria and studied the ecology of marine bacteria in the 1960s, for me it was perfectly rational that interactions among strains and species would be very important, as we're now finding out.
A recent study, to be published in January 2016 in PNAS, on Aeromonas strains that interact to create a pathogenic state, mainly necrotizing fasciitis of a wound, which resulted in the patient undergoing multiple amputations, would not have been possible to understand without the capacity to identify to strain. So that very early work was critical in building the foundation for our rapid, accurate, and actionable diagnostic tools at CosmosID.