Special Report: Entering A New Era In Scientific Computing

When the sixth annual Scientific Computing and Automation Conference and Exposition begins in Philadelphia tomorrow, conferees will be hearing about hundreds of scientific software packages that are now on the market. By contrast, in 1985, when the first conference was held, only a few applications were available for technical users of microcomputers, mostly in the area of data management and statistics. There were no more than a handful of software packages at the dawn of the so-called microcomputer revolution: primarily spreadsheets, word processors, and database managers, with a smattering of communications packages mixed in. Few scientists used microcomputers in their research, because existing software was too limited in both power and functionality. As the general market for microcomputers expanded, though, more specialized software began to appear. Specialized scientific software development has taken two forms. Existing general applications have been extended and customized, and completely new applications have been developed, usually for a specific discipline or field. Most technical software developments in the past half-decade or so have been of the first kind--enhanced or refined versions of already existing packages. Most scientific applications depend on the acquisition, storage, retrieval, and handling of data in some way. Thus, many types of software originally developed for general business applications have been modified to create a user interface that is specific to science. Existing industrial applications for large, mechanized systems have been adapted for instrumentation and control, laboratory automation, and scientific project management. The underlying software in these ad- aptations remains basically the same; only the details of implementation have changed. Take project management. Laboratory project management is not that different from project management for a construction company. One primary difference is that a laboratory management system, such as LabTrack, may allow for data input via bar code or other sample-identification means, as well as via keyboard. Another commonly used genre of programs for both business and science, database management software, works on the same principles regardless of whether the data involved are derived from the office or the lab. However, scientific database management packages may include built-in statistical routines not found in business applications, or may have graphical or data-acquisition capabilities not needed for business purposes. But while many new scientific software packages are essentially new adaptations of existing applications, some programs are entirely new. These programs have no equivalent function in business, industry, or general-application areas. Image processing and molecular modeling programs, for example, are of this variety. While there are similarities between these types of software and more general CAD/CAM (computer-assisted design and computer-assisted manufacturing) programs, such similarities are superficial. Likewise, statistical routines needed for scientific use are not those used in business, even though they may be based on similar mathematical principles. Not surprisingly, applications developed specifically for sophisticated technical use tend to be more expensive than the applications that are essentially reimplementations of more general-purpose software. However, there are many very specialized small programs, largely developed by individuals or small groups, that are generally cheaper than the more commercial packages. Since these applications are not widely advertised, they tend to escape notice. When they can be found they may be very useful, although they suffer somewhat from lack of technical support. Scientific disciplines vary considerably in their emphases and in the types of computer support desirable. Still, there is a great deal of overlap among disciplines. Broadly based scientific software packages include database management programs, spreadsheet-modeling software, statistical and graphical software, technical word processors, data acquisition programs, and instrumentation control software. Following is an examination of several broad disciplinary areas, and the impact that scientific software has had on them. Note that there is some overlap in the various areas, since there is considerable overlap in use of software. For example, many database and statistical applications are used in virtually all disciplinary areas. This designation covers several disciplinary areas, including physiology and psychology, kinesiology, exercise science, and neurobiology. Because of the broad area covered, a wide variety of software packages has become useful during the last decade. Among the more specialized computer applications used in the neurosciences are motion analysis systems (The Scientist, July 9, 1990, page 29). These systems are hardware/software combinations that are used to obtain and analyze data from moving objects, especially the bodies of research subjects. Such data are extremely valuable in some areas of neuroscience research, such as stroke rehabilitation and the study of brain injury. Neuroscience researchers also frequently use more generalized data acquisition and analysis software. Several popular applications on the market, such as ASYST and Lab-Tech Notebook, can be used for acquiring data directly from a research subject or an experimental setup. Such programs also include extensive data storage, handling, and retrieval capabilities; statistical and curve-fitting functions; and graphic displays. Graphic display of data is particularly important in some types of neuroscience research, and many labs find these large, integrated packages ideal for the variety of tasks they handle. A relative newcomer to the field is BrainWave, an integrated data acquisition, analysis, and display package designed to appeal particularly to neuroscience researchers. To some extent, neuroscience interacts with research in expert systems (software that emulates human decision-making processes), with work being done on precisely how information is processed by the brain, and how (or if) such principles can be applied to computer design. While not strictly used in neuroscience research, there is some interaction between designers of neural net software and neurobiologists. Something of a catchall designation, this area is the province of laboratory managers, project managers, and those entrusted with managing the nuts and bolts of research. Less dependent on numerical computation than more "pure" areas of scientific research, this field requires software that deals with instrumentation control, laboratory and industrial automation, and project tracking and management. This is also the only area that deals extensively with expert systems software. There is relatively little difference between scientific project management software and business project management software. Most project trackers, like the popular Who/What/When, or Info-XL, can be--and are--used in laboratory management. However, a recent product, LabTrack, has been introduced specifically for scientific project management. A major claim of this software is that it will accept bar code and other automation-specific input, while most business project managers accept only keyboard or disk-based input. LIMS (Laboratory Information Management Systems) are a larger and more sophisticated type of software, with more capability than project managers. LIMS such as Radian's SAM (for Sample Analysis and Management) System and Perkin-Elmer's CLAS (Chromatography Laboratory Automation System) maintain historical database records and provide for extensive analysis of laboratory operations Moreover, they generally include sophisticated report-generating facilities and some presentation graphics capabilities. In addition to general statistics packages, industrial researchers tend to use a specific type of statistical software known as quality analysis and control software. Quality control software generates specific types of charts and reports, such as pareto charts and n-charts, which assist in the quick detection of anomalies in production or operations. There are an enormous number of quality control packages on the market, including BBN Software's QCA (Quality Control Analyst) and NWA's Quality Analyst. Various journals in the field of quality control regularly compile lists of available quality control software, which should be consulted for the most recent information. R&D scientists usually need good report-generating software. For this, they may use a technical word-processing program, such as T3, Exact, or Expressionist, or they may depend on an integrated package, such as BBN Software's RS/Explore, which provides report generating together with routines for analyzing technical data. Data transmission between remote sites and central facilities is also a concern. Many general-purpose communications packages, such as Close-Up or Carbon Copy, although designed for business use, work well for transmitting scientific data and remote management of scattered computer facilities. There are few, if any, communications packages intended specifically for scientific purposes, since there is generally little need for such applications. There are several expert system development packages available for industrial applications, including Radian's RuleMaster 2 and the less expensive Socrates. It should be noted, however, that many workable expert systems are built without such development tools; an expert system can be designed in any programming language, including FORTRAN and BASIC, although development tools, or "shells," can make the process much faster and easier. Virtually all scientific disciplines have similar needs with regard to database management, telecommunications, technical word processing, and report generation. Statistical analysis is also a universal need, though exactly which routines and procedures are required depends on the disciplinary area. But the fields of microbiology and molecular biology are somewhat anomalous in that they generally require imaging and graphical modeling software. There is some overlap with chemical applications (see the next section) in terms of molecular modeling packages, such as the very large and powerful ChemText system, and much smaller but still useful packages, such as MOLGRAF, Chem3D, and ChemDraw. The rise of recombinant DNA research during the last decade has also produced several software tools designed specifically to assist in such research. Plasmid Artist, by Gene-Systems, is a Macintosh-based system for assistance in gene mapping, and TextCo's DNA Inspector IIe does DNA sequencing, as do DNASIS and The Recombinant Toolkit. Disciplines that depend heavily on image processing, image editing, and graphical displays tend to use more Macintosh-based software than do studies that are heavily data-dependent or that need number-intensive statistical routines. While the dichotomy is shrinking rapidly with recent hardware developments, it is still true that IBM systems have better software for handling very large database applications and number-intensive statistics, and Macintosh systems still have slightly better capabilities for editing and manipulating physical images. WIMP (Wisconsin Interactive Molecule Processor) is another popular molecular editor, as is the Swiss WAALSURF. Such packages are also used by pharmacology researchers, particularly those working in drug design, once more illustrating the broad overlap among scientific disciplines. Chemistry is such a broad area that one can find room for almost any sort of software use within it. While most chemical disciplines can make use of the general-purpose data acquisition and handling packages as well as the scientific statistical applications, chemical fields like pharmacology and pharmaceutical research also may benefit from bibliography management software, such as Pro-Cite, dms4Cite, or RefBase. Those that include modules designed to interact with online databases, such as the National Library of Medicine databases or the Dialog database system are particularly useful for pharmaceutical research, which involves extensive online searches for formulation or patent information. The National Library of Medicine provides its own software, Grateful Med, for searching its own databases, but many researchers prefer to use a bibliography manager with a database-interaction module that will download and standardize formats. Organic chemists may find a molecular modeling package, such as ChemText, useful. ChemText is a very large, powerful application that allows a user to draw atoms and molecules onscreen, and to rearrange elements and bonds at will. This program also interacts with the large molecular database, REACCS, maintained by the vendor that sells ChemText. While virtually all areas of chemistry use statistical programs of one kind or another, chemical researchers are particularly heavy users of curve-fitting packages (The Scientist, April 16, 1990, page 30). Jandel, one of the oldest vendors of scientific software, has several curve-fitting/plotting programs available, including one called PeakFit, which is scheduled for release this month. PeakFit is particularly useful for researchers using spectroscopic data because it includes a number of unusual curve-fitting routines not found in other popular packages. TableCurve, also available through Jandel, is an excellent general-purpose scientific curve-fitting program. One class of software that is generally useful to all scientific disciplines, but is particularly prevalent among chemists, is that of equation solvers. One of the very first scientific/technical software packages marketed, back in 1983, was TK-Solver, which is still available (though substantially updated, and under different ownership). MathCAD, a more recent equation solver, has also achieved considerable popularity. The biomedical sciences are extensive and overlap with many other disciplines. Physiology, immunology, and epidemiology are only a few of the biomedical subdisciplines that may profit from computer technology. Immunologists, for instance, may use many of the same software packages as do chemists. There are also a good many specialized software packages for use by biomedical researchers. For example, Biosoft's AssayZap, originally developed for use in radioimmunoassay, has been expanded in scope to include other types of assay. Numerous clinical and laboratory packages are available for specialized purposes. Nutrition Wizard, for instance, is a nutrition database and planning application for use by dietitians and nutrition researchers. Octopath is a large, multidisciplinary pathology computer system designed to promote efficiency in the running of a pathology department, including modules for histology, cytology, clinical chemistry, microbiology, hematology, and blood bank functions. Cellsoft 2000 Practitioner System (and the larger Cellsoft 3000 system) are hardware/software systems for conducting detailed semen analysis. In addition to the specialized packages mentioned here (and many more), the biomedical disciplines depend on data handling, analysis, and presentation as much as any of the other fields. This is true even though sample sizes may be smaller than those in most chemical applications, with the exception of fields like epidemiology, which is data-intensive and requires the use of large database packages and intensive statistical analysis programs. Some of the most popular database and statistical applications follow. The most important areas of scientific computing have to do with data: acquisition, storage, retrieval, and handling. Data handling affects all disciplines, but sometimes in different ways. Molecular biology may use image processing as a means of acquiring basic data, while pharmaceutical testing processes might use direct input of data from a spectrophotometer or gas chromatograph (via LabTech Notebook, for example, or NWA's Gage Monitor), with no necessity for visualization. Both might use LabTrack for project management in the laboratory environment. Bibliography managers (Pro-Cite, dms4Cite, RefBase, and many others) are specializet forms of data-base management software valuable to all disciplines. All scientific disciplines share certain requirements with regard to data handling, statistical analysis, technical word processing, and graphical presentation. While specific packages in conjunction with particular disciplinary areas have been mentioned, this does not mean that a specific package is intended for use only in that discipline (unless the function of the package plainly limits it to that area). For example, neuroscientists and laboratory managers may find Pro-Cite, a package heavily used by chemists, equally valuable. Several of the most popular packages in the major areas of database and statistical software, as well as some of the other types of software in common use for scientific computing, follow. Database Software: All scientific research rests on data. Consequently, database management software is one of the most important classifications of scientific software. Still, the principles of handling data don't vary much between business and science, which means that general-purpose database software can be used for most scientific applications. Ashton Tate's dBase III+ and the similarly structured FoxBase and Clipper are often used for scientific programming. In addition to the popular relational database packages, such as dBase or Revelation, which will run on the IBM PC/XT/AT as well as on machines built around Intel's 286 and 386 chips, Oracle provides a popular and very powerful relational database package that runs on high-end microcomputer systems and on the larger VAX machines and workstations from Digital Equipment Corp. While these general-purpose database managers can be--and often are--programmed to handle virtually any scientific database application, such programming can be expensive and laborious. Consequently, a number of scientists may opt to use the large integrated packages, such as LabTech Notebook and ASYST. These programs combine data acquisition software with database management and statistical analysis capabilities, and also provide scientific graphics. On the surface, these integrated packages seem to be the ideal answer to scientific computing needs. In practice, however, many laboratories and individual scientists have specialized needs--in terms of statistical analysis, graphic presentation, or data handling--that are better answered by using separate software packages. Many individuals simply have a favorite application that they are reluctant to give up in favor of a larger integrated package that may take time to learn or that may provide unneeded facilities. Statistical Software: There are hundreds of statistical software packages available on the market. Among the most popular of the very large, powerful applications are SPSS/PC+ and SYSTAT, both with wide arrays of statistical procedures and good graphical capabilities. These packages are used in numerous scientific and engineering disciplines. However, such programs are fairly expensive, and require time to learn to use. At the other end of the spectrum is InStat, a pop-up statistical calculator intended for the "quick-and-dirty" statistics that most scientists commonly do. InStat, which is inexpensive, takes no effort to learn and only a few keystrokes to use. It is probably good for more than half the normal statistical calculations done by an individual researcher. Between these extremes lie a number of specialized curve-fitting packages and libraries of scientific statistical routines, such as the Library--Scientific and Engineering Applications and the Library--Statistical Applications, and Jandel's new PeakFit, which includes a number of routines for fitting spectroscopic data that are not available even from the large packages. Other Software: Besides data-base and statistical software, virtually all scientific disciplines can benefit from increased productivity and efficiency by using technical word processors (T3, Exact, Expressionist), mathematical processors (Mathematica, Derive, and the new Theorist), equation processors (MathCAD, PointFive, Eureka, and Dalin's The Scientific Wheel), and administrative tools (Grant Tracker or paired Grant Manager and Personnel Manager packages from Niles & Associates). There are also many plotting and graphics packages, such as Harvard Graphics, CA-Cricket Graph, Jandel's Sigma-Plot, Atlas*Graphics, Statgraphics, Temple Graph, TECH*GRAPH*PAD, and the companion to InStat, Graph-PAD's InPlot. Diana J. Gabaldon is assistant research professor at the Center for Environ- mental Studies, Arizona State University, and editor of Science Software.

Special Report: Entering A New Era In Scientific Computing

Interested in reading more?

Become a Member of