The topic of the presented report is related to cognitive sciences: issues of information perception, molecular genetics, neural networks and modeling of consciousness in general are touched upon. Genometry - a class of information visualization algorithms in the space of biophysical attributes of molecular genetic coding based on the system of Walsh's orthogonal functions. The application of the genometric approach allows to ensure the transition from the semantics of the analyzed signal (as a meaningful set of symbols of a certain alphabet) to semiotics (sign representation, where the final display has a new quality and can be interpreted as a sign). From the position of neurophysiology such semiotic systems can be described by "grandmother's neuron" in the appropriate structure of the neural network. This allows us to speak about a new approach to the study of the phenomenon of consciousness from the position of graph models and molecular genetic algorithms. Evolving random graphs of Erdos Renyl are considered. When a random graph becomes more complex, percolation transition creates large structures and clusters, which can be interpreted as primary elements of consciousness. The manifestation of self-consciousness in the complication of the system of such clusters is also discussed.