I also think that MEME is good for phage display sequences. If you want to test for statistically enriched motifs then you can use SLiMFinder (http://www.southampton.ac.uk/~re1u06/software/packages/slimsuite/). You need to tweak some of the settings as it is designed for whole proteins but the original SLiMFinder paper includes an example (http://www.ncbi.nlm.nih.gov/pubmed/17912346). I am happy to make suggestions and help to get it to work. Unfortunately, the SLiMFinder webserver is currently unable to run with the required settings, so you will need to download the program - or I would be happy to run data locally and send the results back.

If the phage display peptides are already aligned and you want to construct a consensus regular expression from them, you can also use SLiMMaker: http://bioware.soton.ac.uk/slimmaker.html.

Nicholas, a student of mine wrote a program called PhageClust in his master thesis. See this link for the text with illustrations and a zip file with source code and binaries: http://is.muni.cz/th/324635/fi_m/?lang=en

Unfortunately, the text itself is in Czech, but fortunately, there is a readme file (attached) in English that explains how to run the program. I suggest you also contact him directly if you need more information (http://www.recamo.cz/cz/tym/vyzkumny-tym/).

Dear Nicholas,

To generate a consensus from peptide phage sequences, I used MEME Suite:

http://meme.nbcr.net/meme/cgi-bin/meme.cgi

They are few options you can change to tailor your motif search. I had already used to find motifs from few to 2000 sequences and it worked nicely.

I also think that MEME is good for phage display sequences. If you want to test for statistically enriched motifs then you can use SLiMFinder (http://www.southampton.ac.uk/~re1u06/software/packages/slimsuite/). You need to tweak some of the settings as it is designed for whole proteins but the original SLiMFinder paper includes an example (http://www.ncbi.nlm.nih.gov/pubmed/17912346). I am happy to make suggestions and help to get it to work. Unfortunately, the SLiMFinder webserver is currently unable to run with the required settings, so you will need to download the program - or I would be happy to run data locally and send the results back.

If the phage display peptides are already aligned and you want to construct a consensus regular expression from them, you can also use SLiMMaker: http://bioware.soton.ac.uk/slimmaker.html.

Hi Nicholas,

We use MCAFFT, multiple alignment software, http://mafft.cbrc.jp/alignment/software/

Good luck,

Michael FIrer

PS. I've created a blog post with the suggested settings if using SLiMFinder: http://seqsuite.blogspot.com.au/2014/01/using-slimfinder-on-phage-display-data.html. (Someone else was asking about it the other day.)

Hi Nicholas,

one thing that you need to decide is whether to include repetitive sequences or not.

depending on what you want to achieve (consensus sequence, minimal binding motif, critical binding residues, cluster-analysis, sequence-binding relationships ...) . There are many different angles on looking at those sequences.

Dear all,

it's good too see what's available! I'd like to add my 2cents, i.e. my personal favourite: Multalin, available at http://multalin.toulouse.inra.fr/multalin/ Maybe not le dernier cri, but at least good for a quick estimate.

Hi Nicholas ....

How many sequences are we talking about here and what kind of phage display?

Phage display is a extremely noisy process prone to generate garbage or converge on a few identical sequences unless a lot of care is taken and diversity is maintained throughout the rounds of biopanning.

Any alignment software will do really in the initial stages ..... you are just looking generally at what your process is resulting in.

Before diving into mofifs and structures it would be a useful to test if your panel of sequences result in interactions with targets that are meaningful. This means testing your sequences with many irrelevant antigens or targets to measure whether the interaction is specific. Every molecule binds to every other molecule .... it is just a matter of how much. To be meaningful your interactions should be both strong and very specific.

I hope it your targets are all relevant as many of the ones our lab generated were clearly not.

Thank you for all the ideas, I'm going to give a few a try. To answer a couple of questions:

We used next generation sequencing to collect the results, which means that we have _lots_ of sequences, and I'm really trying to find a consensus from the most abundant ones. I can almost do so by eye, but I am sure that software will give a better result.

The library was NEB's PhD-7 system, which adds a 7-mer to the pIII protein (so low copy number).

The plan now is to order a good peptide to test to show that the binding is specific and strong. I'll feedback when I have results...

Just to feedback on this thread:

I tried several of the approaches here, and there were several with different strengths and weaknesses:

MCAFFT worked well, and highlighted (at least by eye) which sequences were similar. It was less useful for more complex thoughts, such as which sequence precisely to choose for follow-up.

MEME was OK, but frustratingly only accepts inputs of at least 9 amino acids. My library was of 7-mers, so this required a little effort to randomly extend the sequences to get any sensible results. The good point of MEME has a very useful output, showing the propensities in the sequences very clearly. Good for publications and presentations.

SLiMFinder worked well too. The good point about this was the very good statistical results, and the excellent documentation to support it (thanks again to Richard). I'm still working through how best to present the results.

I have used a combination of all three approaches to design some peptides for follow-up. Fingers crossed!

Thanks for the feedback, Nicholas. Good luck with the peptides!

Have you tried UniPro Ugene? Its one of the best programs I have used to get consensus sequences including those for proteins and RNA.

MEME is alright. The whole thing depends on the nature of your starting library. Random NNK libraries are huge and tend to give you back a ton of crap in addtion to interesting things. Encoded-sequence libraries are probably better, but a bit harder to pick the motifs out of. The solution here can be to look for amino acids that evolve more slowly than other ones in the isolated fragment.