The proteomic profilings developed from plasma samples of opiate-administered SIV-infected rhesus macaques demonstrate broad suppressive biological responses. Such responses were not observed in our prior study in which monkeys were only infected with SIV (Wiederin et al., 2010
). We posit that this is secondary to broad host homeostatic responses elicited by drug. Indeed, support for this conclusion rests in prior clinical and pathological evaluations of the animals studied. The in vivo
monkey study strongly suggested that opiates can confer protection against the ravages of chronic viral infection in select susceptible animals under these well-defined clinical circumstances (Donahoe et al., 2009
). This stands in contrast to the common notion that the health of HIV infected individuals deteriorates more rapidly when using drug of abuse. (Cherner et al., 2005
; Chana et al., 2006
; Huff 2006
) However, opiate responses may not be uniform. In working to formulate a role of opiates in HIV-SIV/AIDS, the plasma samples from the prior study, due to its controlled conditions, proved most attractive for proteomic investigations. We found experimental conditions in this study as optimal; however, in reality, drug abusers do not have guaranteed access to drugs on a regular schedule and commonly have concurrent infections such HCV and/or malnutrition. We can readily parallel experimental conditions among animals and animal groups relative to housing, food, medical care and other issues that affect disease and health. Interestingly, what was observed showed far greater biological variances than the technology and methodologies combined. It is nevertheless likely that specific signatures for global protein suppression that were found through proteomic profiling are relevant for opiate use. Yet, the source of biological variances still awaits further definition.
We posit that any success realized from proteomic based biomarker discovery utilizing human samples must be carefully designed to acquire meaningful data and to limit effects of co-morbid factors. An important need is the assembly of clinically well-characterized patient groups. In a strictly controlled experiment, such as one using monkeys, four to eight animals can usually generate statistically significant data (Paweletz et al., 2010
). Nonetheless, if other factors are introduced during execution of the experiment, it can lead to variable outcomes as was seen in the current report. We have calculated that the required number of subjects for the current study would be 54 animals (9 animals per group) to obtain a power of 91% to detect the differential protein expression due to different drug treatments (morphine compared to saline) to detect the changes among different time points (baseline as compared to acute or chronic) via two-way analysis of variance. It is likely that similar studies performed with human clinical material will require even larger numbers of subjects. It is now well accepted that quantitative proteomic studies need to be performed using higher number of un-pooled samples to support statistical analysis; however, the exact minimal number of subjects remains unclear (Levin et al., 2009
). In many instances of clinical translational studies, it will be very difficult to achieve this goal as costs of such proteomic profiling will be significantly increased. To tackle this, we envision an experimental approach that will consist of two steps. First is to perform the experiment in an animal model under controlled conditions, and second is to design a proteomic experiment using human material that targets specific protein groups in a defined clinical settings. Statistical analyses applied to such an approach w previously published (Oberg and Vitek 2009
; Cairns 2011
Taken together, results of this study provide an explanation as to why, despite convincing quantitative measures seen during biomarker discovery, protein candidates often fail during subsequent validation phase. We posit that in chronic conditions, biological but not technical variance is a major obstacle in biomarker discoveries. Although by using a well-controlled animal model and reduced technical variability as much as possible at this time, we still observe variable changes in the plasma proteome as a consequence of co-morbid pathologic and exogenous conditions and differences in genetic background. We envision that to overcome such a hurdle better correlates need to be made in improving experimental models, in system biologic statistical and computational approaches and in using better defined and more relevant clinical endpoints.