Skip to main content
Premium Trial:

Request an Annual Quote

Matrix Science Founder John Cottrell Discusses the Business of Mascot

Premium

At A Glance

Name: John Cottrell

Age: 47

Position: Co-founder, Matrix Science, London, since 1998

Prior Experience: Managing Director, Thermo Bioanalysis, UK, 1997-1998

Research Manager, Finnigan MAT, UK, 1987-1997 (renamed Thermo Bioanalysis in 1995)

Development Engineer, Kratos (now Shimadzu), UK, 1982-87

BSc and PhD in chemistry, University College, London

 

How did you get into the mass spectrometry software business?

I started off as a development engineer back in the ‘80s, working with Kratos, which is now part of Shimadzu. We were developing magnetic sector instruments: That was the early days of using mass spectrometry for peptide and protein work. These were not benchtop instruments, these were a roomful of instruments. The main technique in those days was FAB, fast atom bombardment, which has since become obsolete. I worked on a number of different developments aimed at improving the analysis of peptides and proteins by mass spectrometry. Then there was an opportunity to move to Finnigan in the UK because they were setting up a new research and development operation with the intention of developing a benchtop instrument for peptide and protein analysis That seemed like an attractive opportunity, and I moved to Finnigan in 1987. What we developed was the first benchtop MALDI system. It wasn’t the first commercial MALDI instrument — Vestec brought that one out — but what we were trying to put together is a simple benchtop instrument that could be used by non-specialists, and that took a bit longer. It was called the Lasermat.

How did Matrix Science come about? You left Finnigan to start the company?

Yes, with a colleague called David Creasy. We were working together at Finnigan in the MALDI business unit — by this time it was part of Thermo, and in fact, it had been renamed Thermo Bioanalysis — when Thermo decided to close down the UK operation and move the products to Santa Fe in New Mexico. At that time we decided we would like to look for other opportunities; We looked around and saw a very strong opportunity to do something interesting on the software side, and the chance to start a small company.

Where did the technology come from?

We had always had close contact with Darryl Pappin, who at that time was working at the Imperial Cancer Research Fund, ICRF. I had known Darryl for a long time, ever since he got interested in mass spectrometry. We kept in touch and discussed what he was doing, so I knew he had developed some database search software called MOWSE. I knew that ICRF was interested in licensing the software to a commercial company to bring in some license revenue. It just all seemed to fit.

We were able to take the code which Darryl had written as a starting point for a product called Mascot, and get to market much more quickly than if we had started from a blank sheet of paper. We were able to start shipping product a little over a year after we started the company. It was just the two of us for the first two years, but we have grown slowly. We are still not a large company; there are now seven people; six in the UK and one in the US. Matrix Science was initially funded by the founders. Starting a software company is not as expensive as starting a hardware company, so we did not have to get outside investment, which has given us a certain amount of freedom.

How many users does Mascot have, and how do you market it?

A lot of casual users use it for free on the website, and that’s very busy — thousands of people. The number of people who have licensed Mascot for in-house use is in the high hundreds. The larger installations and users tend to be pharma and biotech, but numerically we have large numbers of academic users, who often got the software with an instrument, because one of our distribution channels is to have mass spec manufacturers bundle the software with instruments. Currently we are working with Bruker, Shimadzu, Sciex, and Applied Biosystems, all of whom offer Mascot with their instruments, so we pick up quite a large number of smaller users through that channel. We also sell direct; there are advantages to both channels.

You also distribute through Infocon in Japan?

Because of the distance and the language, it is a challenge to sell directly into Japan, and Infocon has done a wonderful job promoting Mascot in Japan. We have a very strong position now, and we are very pleased with them. They provide a good level of technical support to their own users and have developed a high level of expertise in Mascot.

Doesn’t Mascot compete with software developed by the instrument vendors themselves?

Thermo Finnigan and Waters/Micromass have their own products and would see Mascot as a competitor and so do not handle it. Sciex, for example, has developed a package in-house, but they are also happy to offer Mascot and give the customers a choice. We are happy to compete on the basis of features and technology, and none of our arrangements are exclusive in any way. Bruker has long offered Mascot but I believe they also have a [software] arrangement with Genomic Solutions.

We regard our strength and our niche as being vendor-independent. We aim to offer a solution which is applicable to all the different mass spectrometry products which are popular for protein work.

Is it a problem to convert data from different instrument vendors into a format that Mascot can use?

Most people are using Mascot to search peak lists, not the raw data. They are just text files, and it’s very straightforward to support the different formats. Now, there are limitations to that. Sometimes the peak detection is the weakest link in the chain; if it is not done properly, then that severely limits the quality of the match that you can get from any search engine. It’s certainly our strategy to be able to work directly from the binary files and do our own peak detection, and we have a product which will be launched very shortly and is currently in beta test, which will allow us to do our own peak detection on all of the popular binary file formats. The vendors supply function libraries with the instrument, which allow access to these files.

Do you see value in the proteomics standards initiative HUPO is pursuing?

This is something we very much support, and particularly from our standpoint, it’s really a benefit and an advantage to us whenever there is any kind of standardization. It’s early days: Currently the discussions are about what should be done and how it should be done, but things are moving forward. It would make life easier for the user; they could create an input file which they could then submit to multiple search engines, or they could export search results to a relational database.

How can database search algorithms be improved?

There is an enormous amount to do. We have a very long wish list of new functionality, different ways of doing searches, getting more information, more flexibility, and we just wish we could move forward faster. We are currently bringing out a new release [of Mascot] approximately every six months.

What is highest on your and your users’ wish list?

The access to the binary files, and being able to do our own data reduction and peak picking is critical, and we are almost there with that. Another very active area is taking the results from Mascot and putting them in a relational database, so they can be data-mined more flexibly. At the moment, we produce HTML reports —basically tabular reports — but they are limited, particularly for very large searches such as MudPit.

Many people seem to be interested in new ways to determine how confident you can be of the results in your hit list. Is that also something you are working on?

With Mascot, we aimed to produce a probability-based scoring scheme to which simple statistical tests could be applied. We believe we have that, and that Mascot meets this need. I am sure it could be improved, and there are other things which can be done. It is critical that one can objectively decide if it’s a real match or a false positive, and we believe we have that already.

What is Matrix Science’s secret to success? Other small mass spec software companies have failed …

Our size is more appropriate to the size of the market. This is not a billion dollar market. It’s a few million dollar market at the moment. It is true that people feel greater confidence in a large organization. However, if the market is not of a size to support that large company, that is a false security. If one looks at software companies, particularly scientific software companies, the successful ones tend to be the smaller ones, and when they grow fast or too large, the numbers no longer add up, and that’s when the problems seem to start. Good database search software is not a word processor; it’s not a spreadsheet. There aren’t millions of customers.

What are the plans for the future at Matrix Science?

More of the same, I think. Continue to grow organically. We need to diversify; we are still, essentially, a one-product company, and that is risky. We would like to be more broad-based, so we are looking for ways to diver- sify but without overreaching, spoiling what we have achieved. We want to stick to what we believe we have an edge in, where we have some specialist knowledge. But being independent is just enormous fun.