Weightless neural network array for protein classification

Proteins are classified into superfamilies based on structural or functional similarities. Neural networks have been used before to abstract the properties of protein superfamilies. One approach is to use a single conventional neural network to abstract the properties of different protein superfamilies. Since the number of protein superfamilies is in the thousands, we propose another approach - one network attuned to one protein superfamily. Furthermore, we propose to use weightless neural networks, coupled with Hidden Markov Models (HMM). The advantages of weightless neural networks are: (a) the ability to learn with only one presentation of training patterns - thus improving performance, (b) ease of implementation, and (c) ease of parallelization - thus improving scalability.