SOFTWARE REVIEW art

Bioinformatics
Guide for Evaluating Bioinformatic Software

[Overview] [Why Evaluate?] [Who Will Evaluate?] [How to Evaluate?] [Needs Analysis] [Presentation] [Technical] [Performance]

Reprinted with permission from Drug Discovery Today, Vol. 4, No. 5, May 1999

Reviewed by Steve Bottomley

(Posted June 11, 1999 · Issue 56)


Overview

There are hundreds of bioinformatic software programs available that serve a variety of applications. For someone new to bioinformatics, there is a bewildering array of freeware, shareware, and commercial analytical programs, and it is difficult to make a decision between them. We are mainly influenced by what our peer group uses, but we still need a method to assess, or evaluate, the software to see if it fits our particular purpose. Established workers in bioinformatics also need some way to determine how best a particular software program integrates into the existing infrastructure. Software evaluation is an art, a science, and a business. It can be as casual as a "let's have a play" approach or as formal as a rigorous, structured and systematic investigation. A large or networked organization would usually require detailed and specific criteria for evaluating software and also effective project management of the evaluation process. What follows, however, is a general guide for evaluating "off the shelf" bioinformatics analysis software. Both an individual PC-user or a bioinformatics/IT manager, responsible for a suite of programs over a local or distributed network, may equally use these guidelines.

Why Evaluate?

The first question that needs to be addressed is, Does the software do the job it is supposed to do - is it fit for the purpose intended? Evaluation also helps decisions to be made between competing products; to match software to the specific needs of the individual or organization; to plan for integration within an existing infrastructure; to ensure effectiveness, efficiency, and quality; and, finally, to save time and money.

Who Will Evaluate?

The end user is an obviously important evaluator. However, when organizational issues must be addressed, then the information technology, bioinformatics, and other appropriate departments need to be consulted for the evaluation process. A third party can also be consulted if necessary and this may be used to provide unbiased, independent advice, and thus help to avoid internal politics or vested interests.

Software can also be used to evaluate software - although you'll then have the "chicken and the egg" paradox of trying to evaluate the software which is evaluating the software!

How to Evaluate?

Evaluation can be approached from several different perspectives encompassing scientific, business, or procedural issues. Invariably, however, evaluation is based on a set of criteria, and the software is assessed based on its ability to meet that criteria. The following criteria can be used for scientific evaluation and is subdivided into the four general areas of needs analysis, presentation, technical, and performance sections.

Needs Analysis

Probably the most important first step is to decide what you want to do, and from this develop your criteria. What features do you want? What do you want to achieve? What benefits do you want from the software? What feel and aesthetics do you want? How does it fit within your existing infrastructure? How many users will there be and what is their competency? What do the users need to perform their duties? How do your requirements compare with the vendors' advertised solution?

The next step is to search for candidate software for your evaluation and to gather the necessary background information. Where do you find candidate programs to help you solve your particular needs? Well, you can check out software databases or lists (box 1), vendors (box 2), colleagues, journal articles and advertisements, trade shows, conferences, industry associations, user groups, Internet newsgroups, and Web sites such as BioMedNet. Some of the background information for the software will have to be provided by the vendor or manufacturer or programmer, provided that their inherent vested interest is recognized. Additional third party information can come from any literature reviews and testimonials from previous and existing users.

Box 1. Databases and lists of bioinformatics software

Finally, you need to have some way of scoring the evaluation. For example, a yes/no answer or a quantitative numbering scale for each of the selected criteria. This is to set a standard for evaluation, to enable fair comparison of proposed benefits and features of the program, and to help discriminate between competing products.

Presentation

In terms of presentation, is the software intuitive and user-friendly? Is the interface a command line, Web-based, X Windows, or other graphical user interface (GUI)? How effectively does it use the monitor screen space - is it uncluttered, easily read, and does it make appropriate use of color, graphics, and sound?

Technical

Technical detail can be further subdivided into method, features, and programming.

Method

What algorithm is used and is it appropriate - is it based on old or new knowledge; what are the assumptions inherent in the algorithm, how good is it; is it fast enough; does it do what you expect? Are there any better algorithms? What sort of testing has been done and is this supported in the literature? Is it a "black box" system using proprietary algorithms, or is it an auditable system open to inspection and assessment?

Program Features or Functions

The questions relating to a software's features and functions are numerous, but some general questions (in no particular order) include:

Programming

Further questions relating to programming include:

Performance

The crucial test of any bioinformatic "off the shelf" software is if it makes biological sense. For example, a program may be designed to find patterns, but do the patterns have any biological significance? Does the software do the job you thought it could do? Is it fit for the purpose you intended? Does it give you error-free, accurate, and precise results? Can you trust the results - especially if it is a "black box" system? What sort of quality control has been performed, and does it adhere to your own company's quality control standards? Finally, is a test data set available to determine a software's performance, and is it validated by a third party or from the literature?


Steve Bottomley is an Adjunct Research Fellow at Curtin University of Technology and Managing Director of SciCon Pty Ltd. SciCon specialises in consulting, research, distribution, training and education in biomedical sciences.


Want to see a review of particular software? Send a suggestion.

Previous Beagle Software Reviews
ColorIt! 4.01 for Macintosh
Reviewed by Ellen M. Quardokus (Posted May 28, 1999 · Issue 55)
GraphPad Prism GraphPad Software, Inc.
Reviewed by Ellen M. Quardokus (Posted May 14, 1999 · Issue 54)
Software Survey: What Can We Review for You?
(Posted April 16, 1999 · Issue 52)
Genetwork: Detection of Potential Target Genes
in Silico?
Reviewed by Giovanni Lavorgna, Edoardo Boncinelli, Andreas Wagner, and Thomas Werner
(Posted March 19, 1999 · Issue 50)
MOLE: Molecular Graphics and Computation
for Windows
Reviewed by Xiaolin Liu
(Posted February 19, 1999 · Issue 48)
Nisus Writer 5.1.3
Reviewed by Marie-Claire Daou
(Posted January 22, 1998 · Issue 46)

more