Identification of the most significant amphipathic helix with application to HIV and MHV envelope proteins

Amphipathic helices, which play important roles in protein structure, occur in a wide variety of lengths. Yet existing methods employ fixed window lengths. We present a hierarchical procedure that identifies the Q most significant amphipathic helices regardless of length. Since the observed hydrophobicities are not normally distributed, test statistics usually employed for least-squares regression are inappropriate for assessing statistical significance of amphipathic helices. We show that an adjusted F statistic provides a good test. An application to the envelope protein of HIV finds an unexpected long amphipathic helix in gp41.