Thorough understanding of neural networks is a prerequisite to analyze how information is processed in the brain. For this one has to understand the connectivity between the computational units of the brain, the neurons. Magnetic resonance imaging (MRI) cannot resolve single neurons, and it is hard to trace the entire neural projections of a neuron using electron microscopy (EM). High-resolution light microscopy is therefore the most efficient way to analyze neural networks. To resolve neural projections, only a small subset of neurons should be visualized at one time. However, this makes it impossible to trace connections between neurons. To address this problem, we visualize not only neuronal fibers of selective neurons but also distributions of their input (post) and output (pre) synaptic sites by driving the expression of multiple proteins: a membrane-bound tag and the tags fused with the proteins associated with transmitter receptors and synaptic vesicles. In addition, gross synaptic density of the brain is labeled using an antibody. Because the latter signal is common across samples, images of different neurons that are obtained from different brains can be spatially matched (or registered) to a standard brain template by comparing its signal density distribution. Co-localization between the output synaptic sites of one neuron type and the input synaptic sites of another neuron type is a prerequisite, albeit not the proof, of their synaptic contacts. Using sophisticated expression driver systems of Drosophila, we obtain an array of 3- or 4-channel, very high-resolution, three-dimensionally matched confocal image datasets of diverse neurons to estimate possible connections between different neurons and between different regions of the brain.