The ob/ob mouse: A Case Study

Rules-Based Medicine and Charles River Laboratories compared the phenotypes of the classic ob/ob homozygote mouse with both its wild-type form and with the ob/ob mutant treated with an anti-obesity drug. The protein expression patterns and ensuing results are presented using the powerful OmniViz data mining software.

The ob/ob homozygous mouse is leptin deficient, hyper-insulinemic and becomes extremely obese. Five wild-type lean mice, five ob/ob mice and five ob/ob mice treated with glitazone were bled following a 6-hour fasting period. These plasma were frozen and shipped to Rules-Based Medicine overnight on dry ice. RBM performed the Rodent MAP on the 15 samples and the results are shown in figures 1 - 7.

The result of initial data analysis is shown in Figure 1. These were the nine Rodent MAP analytes that correlative statistical software suggested to be significant. These data were normalized to the lean control animals. Markers of drug efficacy included insulin and VCAM -1. Markers of potential toxicity included haptoglobin, an acute phase reactant and TIMP-1.

Figures 2-7 chronicle the use of the OmniViz software to mine the same data for significant analytes and their expression patterns. The first step was to normalize the data. This was achieved by calculating the standard deviations from the mean for each analyte from each animal. This is shown in Figure 2 where the analytes with standard deviations below the mean were shown in blue and those above the mean were shown in red.

Figure 3 shows the correlation step of the mining process where the movement of each analyte in comparison with the other 87 analytes tested was performed. A blue line connecting two analytes indicated a negative correlation as their values with relation to the mean were moving in opposite directions. A red line indicated a positive correlation.

These lines of correlation were then reduced to pixels as shown in Figure 4. Again, red pixel indicated positive correlation and blue ones indicated negative correlation. The pattern of each animal in the study was then compared using a multi-parameter proximity algorithm and their relativity or proximity to one another was plotted in Figure 5. The closer two animals were to one another, the more similar their analyte expression patterns.

In Figure 5 all 88 analytes were used to determine their relativity. There was good separation between the wild-type animals and the mutants. However, the separation between the untreated and the drug-treated mutants was marginal. We then utilized analysis of variation and a "p" value cutoff of 0.002 to determine the significant biomarkers as shown in Figure 6. These significant markers were then used to re-plot the relativity pattern (Figure 7) and three distinct groups became easily discernable.

The power of the MAP testing approach coupled with the OmniViz data mining software was clearly apparent in this case study. The eleven significant biomarkers can now be described as demonstrating a diagnostic pattern that can be used to discriminate the two phenotypes biochemically. This pattern's utility in drug discovery was demonstrated by the clear separation of untreated and drug treated groups of animals. Possible future applications include the comparison of other anti-obesity drugs to glitazone, comparison of the ob/ob mouse to other obesity models such as the db/db, as well as patterning of biomarkers indicating new avenues of research into target pathways.