Transmembrane Helix Predictor help page

Transmembrane Helix Predictor class includes three different predictors. The first is a Neural Network-based predictor, the others work with values derived from the Kyte-Doolittle scaler. All the tools use the MaxSubSeq "maximum subsequence" algorithm to improve the quality of the results; they differs in the input treatment,

  • HTMR (NN and Profile)
  • Psi Kyte-Doolittle (Kyte-Doolittle scale and Profile)
  • Kyte-Doolittle (Kyte-Doolittle scale and single sequence)
  • Target: a class of Transmembrane proteins, like Seven-helix membrane receptors, Photosynthetic reaction centers, Cytochrome C oxidase-like proteins, F1F0 ATP synthase subunits, Aquaporins, Clc chloride channels, Photosystem I proteins, Calcium ATPase proteins and many other transmembrane proteins with an all-alpha helix structure. It's also possible to detect isolated or locally distributed membrane alpha helices, due to eventual common features or chemical-physical properties carried by this kind of secondary structures.
  • Output: Output is presented graphically in two colors, according to the legend at the bottom of the image, one for the predicted transmembrane alpha helices and the other for the outer-inner membrane protein sequence parts (loops).
    The Neural Network output is represented by two predicted states for each aminoacidic position, one representing the probability to be in a transmembrane alpha helix and the other to be into a loop. The difference between these two predicted probability is represented as a continuous line interpolating each discrete positional value. For the Kyte-Doolittle scale based predictors, the line itself represent a positional (averaged) hydrophobicity value, rescaled with respect to a baseline value.
    In the table, the results for the segments are labelled accordingly to their predicted feature:
    • Internal - for inside membrane surface loops
    • Transmembrane - for membrane spanning segments (alpha-helix)
    • External - for outside membrane surface loops
    The protein membrane orientation derives by application of further computation based on Von Heijne rules.

For more detailed information see: Fariselli P and Casadio R Comput Applic Biosci 12: 41-48 (1996), Kyte J and Doolittle RF J Mol Biol. 157:105-132 (1982), Rost B, Fariselli P, and Casadio R Protein Sci 5: 1704-1718 (1996), Jones DT, Taylor WR and Thornton JM Biochemistry 33: 3038-3049 (1994).