Detection of protein repeats using the Ramanujan Filter Bank

Protein repeats are tandemly repeating segments within an amino acid sequence. They induce several important structural and binding properties on the protein. So far, the most successful detection schemes for such repeats have used computationally expensive techniques such as dynamic programming algorithms, HMMs, and so on. Classical DSP tools such as STFT, unfortunately, perform poorly in the presence of mutations. In this work, a novel technique is proposed based on the recently developed Ramanujan Filter Bank. Fast, accurate, and involving only simple integer computations, its performance is demonstrated on several well-known repeat families.1