Comparing activity analyses for improved accuracy and sensitivity of drug detection

Abstract Activity (or locomotion) can be one of the most sensitive and broadly affected translatable biomarkers of drug or disease. However activity data often have variance heterogeneity and periods with zero activity, and thus is usually not normally distributed giving the possibility of false interpretation of the data. We attempt to address this issue by developing and comparing different analysis techniques. These include transforming the data (square root and ln) as well as determining the probability of activity. In order to comprehensively assess these analysis techniques they are applied to a variety of different activity data sets, which have varying pharmacological manipulation or diurnal cycle state. These analyses indicate that activity data can firstly be improved by a square root transform of the data, which reduces variance heterogeneity. A further improved step is to analyse the “probability of moving”, which is the most sensitive methodology to detect a change in activity. Thus analysis of the powerful non-invasive physiological marker activity and locomotion can be easily and simply modified to improve accuracy and sensitivity in disease or drug detection.