The technologies of computing as applied to science appear to be mature - at least, perhaps, until the Grid makes its presence more widely felt. But there are changes going on within the subject. Across disciplines as widely separated as engineering and life sciences, there have been recent announcements and developments that were surprisingly similar in their significance. If not innovation, then two other 'I' words - interoperability and integration - appear to be key defining features of the present and future direction of scientific computing.
Developers are recognising that none of them can ever become the 'Microsoft' of scientific computing - no one company can hope to sell a single comprehensive software package and tie its customers into that technology alone. Users want to pick and choose - a software tool from Accelrys here, one from Tripos there, perhaps - but above all else, customers expect to be able to use those tools easily.
So the first notable feature of this current landscape is interoperability - companies who are each other's competitors are having to live with the necessity of making their own products compatible with those of their rivals. And hand in hand with this comes an emphasis on workflow, services, and consultancy. Gone are the days of trying to shift shrink-wrapped software and leaving it to the customers to sort out the problems of integrating the software.
Part of this trend was evident in recent announcements from Mathsoft, which produces Mathcad for the management of mathematical (and engineering) calculations. It has formed an alliance and 'technology integration' with SolidWorks. Part of the Dassault Systèmes group, SolidWorks supplies 3D mechanical design software.
Chris Randles, the CEO of Mathsoft, uses one 'I' word - interoperability - to describe the rationale behind the alliance: 'There was an overlap between Mathsoft's and Solidworks' large customer bases and interest among the users in interoperability, and Mathcad is a design as well as an analytic environment.' He points out that the diversity of engineering design work - mechanical, electrical etc. - does not lend itself to any one 'magical' solution. 'We're not a vertically-oriented tool - everyone has to do analysis, modelling and then document it, so we run the gamut.'
Mathsoft has also established its own office in the UK to cater for the needs of 'volume-licensed' customers, focusing on the installation and servicing of large enterprise-wide systems in both private industry and government. Randles points out that developments in the company's products and services require 'closer interaction with our customers and they want to interact with someone local. When customers need to integrate Mathcad with their local package, we now have local people who can help do the integration with them.' The newly-established UK team will service demand for Mathsoft's new Calculation Management Suite, which is designed to help organisations create and manage their quantitative engineering information corporately.
Even within one 'customer', there can be several different groups of users who do not all have the same interests or skills. If scientific computing is to fulfil its function of making science faster and easier at the lab bench (or the engineering drawing board), then the disparate needs of these groups have to be catered for. A really successful piece of software will show these different users that they have interests in common, and it will serve those interests seamlessly and painlessly. Integration goes beyond the technical issue (in computing terms) of interoperability - getting two pieces of software from two different suppliers to talk to each other. It is also a social and managerial issue of getting two different groups of scientists or engineers within one organisation to talk to each other - using computational tools to do so. Sometimes - indeed, often - even though they may be using the same software from a single supplier, integration remains difficult to achieve.
For Mathsoft, the focus on volume-licensed customers, and the concomitant need for local support, is a clear sign of the importance of the second 'I' word featuring in today's scientific computing lexicon - integration. It applies with perhaps even greater force in another area of science, the life sciences. The issue has been exercising Bryan Koontz, senior vice-president and general manager, Discovery Informatics, at Tripos. Pharmaceutical companies are not bringing enough new products to the marketplace but the cost of failure of candidate-drugs is high, so the companies need to be cleverer than they have been in the past about what projects to begin. 'This means that a broader community of scientists is interested in making better decisions about what compounds to follow up,' says Koontz. He conceded that it used to be mainly computational chemists who were interested in the sorts of tools that Tripos could offer, 'but now we're seeing interest from high through-put screening and medicinal chemists.'
Tripos approached its development in a new fashion. 'In the past, it's been assumed that you had to drop the mega-package onto the desktop and the chemists would use it,' Koontz adds. Tripos has been developing a new system, known as Lithium (see Product Update, page 36) - which it saw as a line of communication between medicinal and computational chemists - a way to integrate two groups of users. By providing molecular visualisation on a PC, it allows researchers to view, share, and understand complex molecular data such as protein-ligand crystal structures, docking results, molecular alignments, or other 3D chemical information.
In the past, medicinal chemists had complained that they could not make the compounds whose properties had been predicted by the computational chemists. The computational chemists saw it as a resource problem - they could not fully support all the projects that the company was undertaking, and could not afford to spend their time on trivial 'handle-turning' tasks, when they wanted to validate new scientific methods to solve the problems that their company faced.
According to Koontz, feedback from medicinal chemists was that they were not themselves using computational chemistry, but that they would if it were packaged appropriately and the list of requirements was comparatively short. He says they wanted 'to be able to align active compounds, do a bit of docking, a bit of SAR analysis but did not want the complexities of molecular modelling'.
'Lithium takes the two sets of requirements and says lets provide the environment where the computational chemist builds and validates the methods while the output is captured in a database and browsed by the medicinal chemists.'
The key to integration is an understanding of the motivations and dynamics of the users. Koontz remarks that 'most medicinal chemists and their managers say it must be quick and efficient to use, so they can get back into the lab to make the compound'.
Koontz has his own personal experience of integration to draw upon, in meeting such challenges. Until late last year, he was CEO of Optive Research, when it was acquired by Tripos. Optive was founded in 2002 by Dr Robert Pearlman, from the University of Texas at Austin, as a technology spin-out from the university. Before the take-over, Tripos was already a distributor of five of Optive's software products, including DiverseSolutions, StereoPlex, and Concord, a program for creating three-dimensional chemical models. Optive's software too was designed for both computational chemists and wet-lab scientists. 'Most of the heavy lifting on product integration is underway, if not completed,' he says. He expects the announcement of new products, aimed at both computational chemists and bench researchers, later in the year because the two companies had complementary rather than duplicating technologies. Some 12 of the 15 Optive software products enhance Tripos' other major software package, SYBYL.
The other 'big beast' in this area of life-sciences computing, Accelrys, also used the 'I' word when announcing its 2005 product launch. In addition to the individual enhancements and new products, the company announced its 'Open Integration Platform', linking its own software and that of its recently acquired SciTegic subsidiary.
SciTegic's main product is Pipeline Pilot, a data analysis and mining system for drug discovery. Spotfire, Accelrys itself, and The BioTeam have recently joined the SciTegic's 'independent software vendor (ISV) partner programme'. Spotfire is developing an interface to launch Pipeline Pilot data processing workflows from within DecisionSite, its visual data analytics application. Accelrys software users will be able to add more information about their data (e.g., derived chemical properties) through the execution of data processing routines via Pipeline Pilot. Through its professional services consulting, The BioTeam will be able to help life science researchers develop infrastructures that integrate Pipeline Pilot. These industry leaders join the six other software vendors that have recently joined the ISV programme.
Accelrys is integrating several applications in chemical informatics and visualisation, structure-based design, and bioinformatics. One of the first will allow users of DS Accord for Excel to call computational protocols within Pipeline Pilot. 'SciTegic recently launched its ISV programme to provide a level-playing field for software vendors that want to integrate their tools in a data pipelining environment. We provide technical assistance to help our partners achieve high quality integration,' says Matt Hahn, vice-president and general manager of SciTegic.
Creative ideas about bringing users together are not the sole prerogative of the larger scientific computing companies. InforSense, a company that spun out of London University's Imperial College in 1999, sees its 'workflow' software as going beyond data integration. According to Judith Bandy, director of external communications, it is 'more a way of combining expertise, existing services and tools, computational capability along with that data, to enable the scientist to be creative and the organisation to benefit from that creative knowledge'.
The company's DiscoveryWorkflow technology 'allows you to capture the expertise of users in the scientific community,' she continues. 'Someone skilled in library design, for example, can create a process that encapsulates best practice and then deploy it across an organisation, so you're sharing the expertise.' In contrast to the rigid procedures necessary in QA/QC laboratories - especially in regulated industries such as pharmaceuticals - which are given expression in Laboratory Information Management Systems (LIMS), the development of a workflow is more adapted to the fluidity of the discovery laboratory. 'It's a very ad hoc process to compose a workflow. Once created, it can serve as a template for others, or they can modify it. We can track who makes changes or what data or tools have been used, so a workflow can be used as a standard process or it can be developed and expanded upon by others. It's not rigid.' Dr Bandy stresses.
InforSense has established an Open Workflow Partner Network for scientific and business informatics to provide interoperability (that word again!) between proprietary data sources and analytic and reporting tools.
Scientists and analysts can still use their chosen applications, but integrated via InforSense's open infrastructure. Partners include Daylight, LION bioscience, Spotfire, and Elsevier MDL.
A significant indicator of the power of the InforSense approach is that, despite its link with SciTegic, The BioTeam has also formed a partnership with InforSense to integrate its informatics clustering kit, iNquiry, with InforSense KDE. This is a key step in offering high-performance Grid computing for bioinformatics.
As remarked at the beginning of this article, technical innovation - in the form of the Grid - does not appear to have percolated through to commercial scientific computing. It may be that interoperability and integration are the necessary precursors and that, this time, innovation will follow rather than lead.
