Structures and science
How did the company begin?
ACD labs began not as a commercial company but as an R&D project by a group of scientists from Moscow State University in Russia. The aim was to predict NMR spectra and Log P - a physical property of compounds. It turned in to a promising technology and then in 1993, an office was opened in Toronto as a North American base for commercial operations. The entire R&D operation - 65-70 people - is still in Moscow, as is the scientific leadership for each product. Sales, marketing, and customer support are in Toronto. All these activities are glued together by a team of product managers, based in Toronto, who collaborate with industry experts and set overall product direction. We are a private company and we grow by pouring our own revenues into R&D.
The product line includes naming chemical structures, prediction of physical properties, and analytical chemistry informatics. In this latter grouping, prediction of NMR spectra is the beginning of what we do: we can compare predicted to experimental spectra and automate the analysis of spectra by attaching structures to spectra. This increases throughput in the labs, and nowadays it's 'high throughput' everything.
How did you begin with ACD?
I joined the company two years ago from Abbott Laboratories, where I was dealing with the management of laboratory information. I sent out requests to cheminformatics vendors for a new system to manage biology assay data and was impressed when ACD said this wasn't what they did. ACD's specific focus is the naming of compounds; the prediction of physical properties and analytical chemistry information. It's not a space we own uniquely, but what we don't find is a company that has the breadth that we have in terms of analytical techniques, and a deep understanding of chemical structure and how that connects with their spectra. Not many other packages predict 95 per cent of the compounds thrown at them and get them right.
What software does ACD provide?
We have 50 products but it's more that we have one product with 50 packages or modules. It allows us to innovate rapidly - if we need new database functionality for one thing, then it shows up in all the packages. We have the same user interface design for all users - you can switch from NMR to mass spectroscopy and only need to learn the MS specific functionality. The rest is the same.
What we're focused on at present is to use our system but 'flipped on its head' for a different problem in analytical chemistry: where you have an analyte but don't know what it is, for example in trying to identify a metabolite. You may have some clues to its structure but you don't know. We're getting good success here.
In a different area, chromatography, one of the biggest challenges is that you don't know where to start in developing a separation protocol: you use general knowledge and optimise from a crude beginning.
Our software can store separations in the database with the chemical structure assigned so that when developing a new method, you can query the database, ask when you have done something similar and then, in silico, put in the new structures.
What about the future?
All the problems we address are happening in a single lab on a single site in a single company: for example, the NMR lab. But at another site there is another NMR lab in the same company. They may be doing the same thing. All the information and data can cross-fertilise - it needs to be enterprise wide. This leads to Web-based viewing tools and Oracle-based data management.
The other strand is that we're connecting up with data used to analyse a sample late in the product development cycle: for example, you may test for impurities in a tablet. Earlier on you'd have validated the chemical structure of the active ingredients and their metabolites. All that data could be re-used late in the cycle. So there's an opportunity to feed that information forward in time. This is a shift on the organisational level, not within one lab, but linking information across organisational boundaries within a big, big company.
A classical LIMS vendor can address this at the level of sample ID numbers, but we are connecting the nature of the sample with its spectral signatures throughout product development.
We are not into custom software; we are a product company: build it once and build it well, then others will benefit. We are here to stay and grow, not like a weed, but like an oak tree.