We present the first tomographic reconstruction algorithm for optical molecular imaging that is based on the equation of radiative transfer. The reconstruction code recovers the spatial distribution of fluorescent sources in highly scattering biological tissue. An objective function, which describes the discrepancy of measured near-infrared light with predicted numerical data on the tissue surface, is iteratively minimized to find a solution of the inverse source problem. At each iteration step the predicted data are calculated by a forward model for light propagation based on the equation of radiative transfer. The unknown source distribution is updated along a search direction that is provided by an adjoint differentiation technique.