The official website of the HHMI Science Education Alliance-Phage Hunters Advancing Genomics and Evolutionary Science program.

Welcome to the forums at Please feel free to ask any questions related to the SEA-PHAGES program. Any logged-in user may post new topics and reply to existing topics. If you'd like to see a new forum created, please contact us using our form or email us at

Membrane proteins

| posted 11 Mar, 2021 17:53
My students love calling membrane proteins this semester for some reason. What do I advise them to do with proteins that have two or more good TM domains (using at least 2 different programs) but no phagesdb BLAST membrane protein calls. So far, I've been explaining that hydrophobic helices can be internal to proteins as well as serve as TM domains. I've told them they need more evidence than 2 or more good TM domains predicted to make a membrane protein call. Am I giving them good advice or should I be more liberal in calling membrane protein?
| posted 12 Mar, 2021 18:47
Your question about false positives is an interesting one. I had always assumed the algorithms were specifically designed to distinguish the differences between membrane domains and simple hydrophobic helices. So I went back to the 2001 paper for TMHMM (doi:10.1006/jmbi.2000.4315). The intro in the paper has a really good discussion on the early methods used to distinguish just that issue. There is also a whole section of the paper on this issue. Bottom line is there are other structures and length requirements that help in the determination which helps distinguish a "real" transmembrane domain. Might even be worth pointing out this paper to students who are interested, if only to read the intro.

As for the issue of false positives with TM-HMM, according to the paper, the algorithm has a specificity of around 99% if there is not a leader peptide, so I think the protocol as defined is a pretty good one and further support from BLAST is not required.

But this issue that a leader peptide reduces the quality of the results is very interesting. Maybe the SEA-phages protocol should be amended if a leader peptide is predicted. That is really a good question for a faculty workshop I think.
Edited 12 Mar, 2021 18:52
| posted 12 Mar, 2021 18:49
Thanks Chris! I will definitely work through that paper with my students.
| posted 01 Aug, 2022 16:26
Could someone remind me why 'membrane protein' seems to be missing in the Pham 'notes' section, and isn't seen as a predicted function when looking at a Phage gene list? The function is in the published Genbank files however. An example below is with Asa16.


Pham 36850 - No 'notes' with 'membrane protein' function?
Asa16_CDS_23 Membrane Protein.

From the Genbank file:

gene 19751..20182
CDS 19751..20182
/product="membrane protein"
| posted 01 Aug, 2022 16:43
Hi Nick,
Membrane proteins, especially those containing 1 transmembrane proteins, are tricky. Very much is dependent on how good the program used to call it is. We are formulating a plan to improve how this work is done, but it is still in process. In the meantime, here is the status of reporting transmembrane proteins.
As it turns out, as Travis revamped the database that informs phamerator, he removed certain products because they were problematic. These include hypothetical, gp, orf, putative, and many more. Membrane protein was included on the list because of the 'iffy' nature of how membrane proteins are called. So what you describe is exactly correct. and unfortunately, a bit confusing. (Today we would make a different decision about this, but that is what we have for now.)
Call membrane proteins to the best of your ability. Hopefully more news to follow soon.
| posted 01 Aug, 2022 18:50
Thanks Debbie! I thought maybe I had missed a new directive.

We have been following the guidance that if there is a single predicted TMD, then it must be called by at least two different TMD prediction programs.

| posted 02 Aug, 2022 01:04
Yep. You got it!
| posted 09 Feb, 2023 23:07
Hi Debbie et al.,

In the January 2023 faculty meeting you updated guidance on TMD prediction. As of now we are using just DeepTMHMM and looking for at least one TMD, about ~17-22aa long.

My question is…how close does that "~" need to be? If DeepTMHMM predicts a TMD of 15 or 16aa, are those close enough? Or does it need to fall WITHIN 17-22aa long?

| posted 13 Feb, 2023 13:45
This answer is all about context. I don't know the answer with the information provided. Look at the whole genome. What makes sense? Now that we know that N terminal hits can be signal proteins, i think where in the gene the membrane protein is found impacts my thinking. Just because we moved to Deep THHMM doesn't mean you can't use other tools to continue your evaluation. In general, keep it simple. Membrane proteins are not functions. And when Phamerator gets this in place, the data is not lost.
| posted 10 Apr, 2023 07:19
I am a little confused with our current calling of membrane proteins. Should we be calling membrane proteins as functions within our annotations? From PhagesDB and Phamerator it looks like there is a move away from that.
Login to post a reply.