First of all True Discovery is a rubric for a statistically defensible and honest assessment of experimental results that takes into account the special nature of genomic data. True Discovery is statistically defensible because we stay within the mainstream of biostatistical methods most commonly used within clinical and basic research with an emphasis on assessments of statistical significance and hypothesis testing.
While we value the broader range of statistical and data mining techniques that have been brought to bear on the analysis of genomic data, we believe that regression based techniques with their emphasis on hypothesis testing and reproducibility will have the greatest utility in biomedical studies. True Discovery is an honest assessment of experimental results because, like traditional statistical treatments of type I and type II errors, True Discovery admits and quantifies the simple truth that some of our leads will be the consequence of random variation in the data.
True Discovery is specifically adapted to the needs of genomic research because we have built the need for normalization and heterogeneity corrections into our regression framework. Rather than separating normalization and analysis, we've built them into a unified regression model. Combining normalization and modeling into a single regression framework allows us to assess the significance of results while simultaneously taking into account the need for normalization of genomic data.
An ad hoc normalization scheme requires evidence that normalization has not modified the significance of our results. By contrast, our regression framework includes normalization so that we are guaranteed a meaningful assessment of the sample variability of both regression coefficients and normalization.