Erik L.L. Sonnhammer, Gunnar von Heijne, and Anders Krogh:
A hidden Markov model for predicting transmembrane helices in protein sequences.
In Proc. of Sixth Int. Conf. on Intelligent Systems for Molecular Biology, p 175-182
Ed J. Glasgow, T. Littlejohn, F. Major, R. Lathrop, D. Sankoff, and C. Sensen
Menlo Park, CA: AAAI Press, 1998
Download compressed postscript file
Download pdf file
Press here to see other material (model, training data, etc). .
Version 2 is very similar to version one, but it builds on a new model,
so predictions are not identical. The web server has been improved (hopefully)
a little. A paper is submitted describing TMHMM in more detail (publication
details not available yet).
This is an example (one protein):
>5H2A_CRIGR you can have comments after the ID
Here is an example:
# COX2_BACSU Length: 278
# COX2_BACSU Number of predicted TMHs: 3
# COX2_BACSU Exp number of AAs in TMHs: 68.6888999999999
# COX2_BACSU Exp number, first 60 AAs: 39.8875
# COX2_BACSU Total prob of N-in: 0.99950
# COX2_BACSU POSSIBLE N-term signal sequence
COX2_BACSU TMHMM2.0 inside 1 6
COX2_BACSU TMHMM2.0 TMhelix 7 29
COX2_BACSU TMHMM2.0 outside 30 43
COX2_BACSU TMHMM2.0 TMhelix 44 66
COX2_BACSU TMHMM2.0 inside 67 86
COX2_BACSU TMHMM2.0 TMhelix 87 109
COX2_BACSU TMHMM2.0 outside 110 278
If the whole sequence is labeled as inside or outside, the prediction
is that it contains no membrane
helices. It is probably not wise to interpret it as a prediction of location. The prediction gives the most probable location and orientation of transmembrane helices in the sequence. It is found by an algorithm called N-best (or 1-best in this case) that sums over all paths through the model with the same location and direction of the helices.
The first few lines gives some statistics:
Length: the length of the protein sequence. Number of predicted TMHs: The number of predicted transmembrane helices. Exp number of AAs in TMHs: The expected number of amino acids intransmembrane helices. If this number is larger than 18 it is very likely to be a transmembrane protein (OR have a signal peptide). Exp number, first 60 AAs: The expected number of amino acids in transmembrane helices in the first 60 amino acids of the protein. If this number more than a few, you should be warned that a predicted transmembrane helix in the N-term could be a signal peptide. Total prob of N-in: The total probability that the N-term is on the cytoplasmic side of the membrane. POSSIBLE N-term signal sequence: a warning that is produced when "Exp number, first 60 AAs" is larger than 10.
At the top of the plot (between 1 and 1.2) the N-best prediction is shown.
The plot is obtained by calculating the total probability that a residue sits in helix, inside, or outside summed over all possible paths through the model. Sometimes it seems like the plot and the prediction are contradictory, but that is because the plot shows probabilities for each residue, whereas the prediction is the over-all most probable structure. Therefore the plot should be seen as a complementary source of information.
Below the plot there are links to
For the example above the short output would be (except that it would be on one line):
"len=": the length of the protein sequence. "ExpAA=": The expected number of amino acids intransmembrane helices (see above). "First60=": The expected number of amino acids in transmembrane helices in the first 60 amino acids of the protein (see above). "PredHel=": The number of predicted transmembrane helices by N-best. "Topology=": The topology predicted by N-best.
The topology is given as the position of the transmembrane helices separated
by 'i' if the loop is on the inside or 'o' if it is on the outside. The
above example 'i7-29o44-66i87-109o' means that it starts on the inside,
has a predicted TMH at position 7 to 29, the outside, then a TMH at
position 44-66 etc.
One of the most common mistakes by the program is to reverse the direction of proteins with one TM segment.
Do not use the program to predict whether a non-membrane protein is
cytoplasmic or not.