Data Tracking Systems - Biomedical Informatics in Translational Research - page 127

analysis should be captured by the data tracking system. The logic why a data track-
ing system should not support data analysis, is that it is supposed to track the exper-
imental data generated in the lab, with an audit trail such that any modifications of
the captured data are tracked. If the experimental data are analyzed and tracked in
the data tracking system, the recursive trial-and-failure process will definitely pose a
major challenge to the audit trail. For a heavily used system, it is certainly not desir-
able to have data analysis compete for CPU time and BUS traffic.

7.4 Ways to Establish a Data Tracking System

So far we have discussed the requirements for a data tracking system. Next we
explore how a data tracking system is established in a specific biomedical informa-
tics research organization. Three approaches are possible: (1) Buy the system off the
shelf, (2) develop the system from scratch, or (3) pursue a hybrid mode for which
part of the system is purchased off the shelf and the rest is developed in house or
with a partner.
Before making an implementation decision, one should first assess the specific
needs of the organization, and then compare these needs to the capabilities of com-
mercially available systems. The requirements outlined in the previous two sections
should be of value to this assessment process. For example, if an organization
focuses mainly on biomarker discovery using a gene expression microarray experi-
mental platform or array-based genotyping platform, then there are several com-
mercial systems from which to choose, and the best way to establish a data tracking
system for such a need may well be to buy one off the shelf. On the other hand, if the
organization uses the protein antibody array as the main experimental platform,
then there is no system currently available on the market (although this may change
with time), so the only way to satisfy this need is to develop one. If the organization
operates on several data platforms, and the available commercial data tracking sys-
tems can satisfy some but not all the needs, then the organization may opt for a
hybrid solution—to buy whatever is available and develop the rest.

7.4.1 Buy a System Off the Shelf
Although peer-reviewed papers and trade magazine articles are available on how to
select a LIMS vendor and test or validate a LIMS product, the publications are in
general either not up to date or focused on analytical chemistry or quality assurance
but not on biomedical informatics research [7-9, 14, 15, 17-19]. Given this limita-
tion in the literature, we suggest some criteria for identifying a commercial data
tracking system for biomedical informatics research. These factors can be catego-
rized as front-end technical features (factors 1-5), back-end technical structure fea-
tures (factors 6-8), and general issues (factors 9-13) and they are discussed next:

1. Is the workflow interface user friendly? A cumbersome interface will
discourage users from using the data tracking system. Having to work with a
cumbersome interface daily will also negatively affect the performance of
those working in the lab.