Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Persistence and Availability of Web Services in Computational Biology

Figure 1

Charts for many criteria used to evaluate the services, by year of publication.

This figure shows numerical values and sparklines [27] for the criteria of every year the NAR Web Server Issue was published, listing explicitly the values for 2003, 2009, the minimum value of these years in blue (labeled min), the arithmetic mean in gray (labeled mean) as a straight line, and the maximum number in red (labeled max). (A) Services analyzed lists the absolute number of services we extracted from that issue of NAR. Some publications describe a whole collection of services, which were not evaluated individually, but rather by criteria apt for collections, and appear as a single service in the graph. (B) Web address reachable is a relative number of URLs that did not return an error message when accessed in a browser (as described in [3]), but may contain services that are not operational while still displaying their regular Web page. (C)–(G) We tried to locate version information, “last updated” information, contact information, example data, and online help/manual, on the service's Web page to the best of our abilities. As the requirements for submissions to NAR Web Server Issues tightened, we see an increase in these numbers, except for the version information, which becomes pertinent as the Web service ages. (H) For services to give us a fair testing possibility, we required either easily obtainable example data or standard file types such as FASTA, PDB, GFF, etc.(cf. Methods). (I)–(J) The percentage reported in service operation verified is taken from the total number of services in that issue. We assigned usability scores from zero to three, services with high usability score either two or three. A high score is assigned to services with clearly arranged user interface widgets, the presence of default values and easily accessible help and usage information. It is low for services with strong restrictions on input data and crowded, unclear user interfaces without documentation. (K) The LT-Score is calculated for every service, on a range from zero to 41, and zero to twelve for every collection of services (see Methods). (L)–(M) As the LT-Score for services with high usability and low usability show, services with user interfaces that are well-arranged also have an above average LT-Score, and services with unclear interfaces score significantly lower. (N) The LT-Score for collections is quite constant over all issues. (O) The number of authors that participated in the survey is reflected in Author Survey Replies. Not surprisingly, authors from more recent services are more likely to respond, not least because their email addresses listed in the publication are still current. The higher number of replies for 2003 reflects the care and commitment the services from this issue have received.

Figure 1