Summary: Autism is a heterogeneous disorder involving complex mechanisms and systems occurring at diverse times. Because an individual child with autism may have only a subset of all possible abnormalities at a specific time, it may be challenging to identify beneficial effects of an intervention in double-blind, randomized, controlled trials, which compare the mean responses to treatments. Beneficial effects in a small subset of children may be obscured by the lack of effect in the majority. We review the evidence for several potential model systems of biochemical abnormalities that may contribute to the etiology of autism, we describe potential biomarkers or treatment targets for each of these abnormalities, and we provide illustrative treatment trials using this methodology. Potential model systems include immune over and under reactivity, inflammation, oxidative stress, free fatty acid metabolism, mitochondrial dysfunction, and excitotoxicity. Including potential biomarkers and targeted treatments in clinical trials for autism provides a potential method for limiting the heterogeneity of enrolled subjects, which may improve the power of studies to identify beneficial effects of treatments while also improving the understanding of the disease.