Abstract Motivation Computational studies of molecular evolution are usually performed from a multiple alignment of homologous sequences, on which sequences resulting from a common ancestor are aligned so that equivalent residues are placed in the same position. Residues frequency patterns of a full alignment or from a subset of its sequences can be highly useful for suggesting positions under selection. Most methods mapping co-evolving or specificity determinant sites are focused on positions, however, they do not consider the case for residues that are specificity determinants in one subclass, but variable in others. In addition, many methods are impractical for very large alignments, such as those obtained from Pfam, or require a priori information of the subclasses to be analyzed. Results In this paper we apply the complex networks theory, widely used to analyze co-affiliation systems in the social and ecological contexts, to map groups of functional related residues. This methodology was initially evaluated in simulated environments and then applied to four different protein families datasets, in which several specificity determinant sets and functional motifs were successfully detected. Availability and implementation The algorithms and datasets used in the development of this project are available on http://www.biocomp.icb.ufmg.br/biocomp/software-and-databases/networkstats/. Supplementary information Supplementary data are available at Bioinformatics online.