The functions of human brains highly depend on the precise temporal regulation of gene expression, and the temporal brain transcriptome profile across lifespan has been observed. The substantial transcriptome alteration in neural disorders like autism has also been observed and is thought to be important for the pathology. While the cell type composition is known to be variable in brains, it remains unclear how it contributes to the temporal and pathological transcriptome changes in brains. Here, we applied a transcriptome deconvolution procedure to an age series RNA-seq dataset of healthy and autism samples, to quantify the contribution of cell type composition in shaping the temporal and autism pathological transcriptome in human brains. We estimated that composition change was the primary factor of both types of transcriptome changes. On the other hand, genes with substantial composition-independent expression changes were also observed in both cases. Those temporal and autism pathological composition-independent changes, many of which are related to synaptic functions, indicate the important intracellular regulatory changes in human brains in both processes.