Hi all,

There are plenty of tools to detect functional divergence using sequence information only. They use conservation measures (i.e. entropy) or evolutionary models (i.e. substitution matrix).

There is a list in this review (not OpenAccess, but ask me for a copy):

http://www.biochemj.org/bj/449/bj4490581.htm

Here is the list (some are using 3D structure information):

Tools URL

3DSim http://3dsim.bioinfo.cnio.es/

Badasp http://www.southampton.ac.uk/~re1u06/software/badasp/

CAPS http://bioinf.gen.tcd.ie/caps/

CHAIN http://www.chain.umaryland.edu/

CheckCov (available from request to the authors)

ConSurf http://consurf.tau.ac.il/

Diverge http://xungulab.com/software.html

Evolutionary Trace http://mammoth.bcm.tmc.edu/

FastML http://fastml.tau.ac.il/index.html

FunDi http://rogerlab.biochemistryandmolecularbiology.dal.ca/fundi.php

GroupSim http://compbio.cs.princeton.edu/specificity/

HMMDiverge http://info.mcmaster.ca/yifei/software/HMMDiverge.html

HyPhy http://www.hyphy.org

Meme http://www.hyphy.org/w/index.php/MEME

MultiRELIEF http://www.ibi.vu.nl/programs/multirelief/

Mutationassessor http://mutationassessor.org/

PAML (CodeML) http://abacus.gene.ucl.ac.uk/software/paml.html

Procov http://www.mathstat.dal.ca/~hcwang/Procov/

Rateshift http://www.daimi.au.dk/~compbio/rateshift/

Rate4Site http://www.tau.ac.il/~itaymay/cp/rate4site.html

RAte Shift EstimatoR (RASER) http://www.tau.ac.il/~penn/raser.html

SAAPdb http://www.bioinf.org.uk/saap/

SDPpred http://bioinf.fbb.msu.ru/SDPpred/

Selecton http://selecton.tau.ac.il/

Shift-Finder http://sites.univ-provence.fr/evol/softwares/shiftFinder/SHIFT-FINDER.htm

SIFT http://sift.jcvi.org/

SPEER http://www.hpppi.iicb.res.in/ss/

SPEL ftp://iole.swmed.edu/pub/SPEL/

TDG09 http://www.homepages.ucl.ac.uk/~ucbtaut/

TestNH http://biopp.univ-montp2.fr/forge/testnh

Xdet http://pdg.cnb.uam.es/pazos/Xdet/

Evolutionary Databases URL

Cube-DB http://epsf.bmad.bii.a-star.edu.sg/cube/db/html/home.html

FunShift http://funshift.sbc.su.se/

FunTree http://www.ebi.ac.uk/thornton-srv/databases/FunTree/

Human PAML browser http://mendel.gene.cwru.edu/adamslab/cgi-bin/paml/pbrowser.py

SDR http://paradox.harvard.edu/sdr

Selectome http://selectome.unil.ch/

The Adaptive Evolution Database (TAED) http://www.wyomingbioinformatics.org/TAED/

In a previous study (in OpenAccess), I used BADASP and CheckCov:

http://mbe.oxfordjournals.org/content/27/11/2618.short

I have also made a tutorial on how to use some of these tools:

http://www.cathdb.info/wiki/doku/?id=tutorials:eccb_t2_studer

Of course, all the tools are making prediction. So you cannot say for sure which amino acids are responsible for functional divergence or not. But that can help to select a few ones for mutagenesis for exemple.

Don't hesitate to contact if you need any help!

Best regards,

Romain

--

Romain Studer

EMBL-EBI

Wellcome Trust Genome Campus

Hinxton

Cambridgeshire

CB10 1SD, UK

Twitter: @RomainStuder

I think you just have to align as many sequences of your paralogues as possible, find conservative regions and see which amino acids are different. If you know the residues belonging to the active/binding sites, have a specific look into these parts of the sequences.

In addition to my phylogenetic analysis I wanted to identify such regions. Because the program that was used for this analysis requires a lot of CPU usage, I only punt in 15 homologies of both genes. So, I have the alignment and a phylogenetic tree, both in correct format, but now comes the tricky part. DIVERGE requires clustering to calculate TYPE I functional divergence regions. I have tried clustering and got results, but I am not sure if I have created the right clusters, and don't undestand the results.

I often get to the problem you describe re:BI approaches. I did not use DIVERGE, but cannot you do something like we did in:

Structure-function relationships in the 2-oxo acid dehydrogenase ...

www.ncbi.nlm.nih.gov/pubmed/18004749

In deed, it's an interesting problem to work on. I suggest you to initiate with 'Bioinformatics' study followed by 'Reagent Based laboratory Work' This order would certainly be providing you certain interesting and fascinating data.

A good way of finding functional regions is checking their evolutionary trace, as Victoria said. I use Consurf in combination with Chimera (for visualiztion of results).

Dear Victoria, both genes have no 3d structure. Any other suggestions?

Dear Antonio, I have used conserv but how can I identify/calculate the regions of interest? The program just gives conservation scores. Is there an additional feature?

In addition to the excellent above suggestions I would recommend considering your candidate amino acids physico-chemically. As not all amino acids are equally similar some substitutions will have a bigger effect on structure/ function than others. For example a small hydrophobic amino acid substituted with a large hydrophillic may have a larger effect than a like-for-like substitution in a key structural region. There are several ways to consider amino acid similarity, we recently published a consensus/review at http://link.springer.com/article/10.1007%2Fs00239-013-9565-0

http://link.springer.com/article/10.1007%2Fs00239-013-9565-0

There are some papers published that can help you. The first two are available on line (free).

Protein Sci. 2004 Jan;13(1):25-31.

Fygenson DK, Needleman DJ, Sneppen K.. Variability-based sequence alignment identifies residues responsible for functional differences in alpha and beta tubulin

Mol Cell Biol. 1994 Aug;14(8):5066-75.

Functional differences between HOX proteins conferred by two residues in the homeodomain N-terminal arm.

Phelan ML, Sadoul R, Featherstone MS.

Mol Ecol. 2005 Jun;14(7):2223-31.

Activity variation in alcohol dehydrogenase paralogs is associated with adaptation to cactus host use in cactophilic Drosophila.

Matzkin LM..

De ar Rene,

Obviously it matters what type of proteins you are studying (membrane bound, cytosolic) and what you want to get out of it (ligand binding, signaling).

However, we have had great success using proteochemometric modeling. We applied it to the adenosine receptors (all human paralogs AND all rat paralogs) and the results were published here : http://pubs.acs.org/doi/abs/10.1021/jm3003069

While we constricted ourselves to the binding site, this can easily also be applied to whole sequences.

We applied the same technique to enzymes, transporter proteins, etc. Please see my personal website for further publications : http://www.gjpvanwesten.nl/publications

Cheers!

Addendum: the major advantage of the statistical proteochemometrics approach is that it does not need structural information (while the availability of 3d info can improve models).

You can use the Protparam tool that computes the physicochemical paraters . It will give you the amino acid composition of each paralog. You can then construct a table for comparative studies.

Hi all,

There are plenty of tools to detect functional divergence using sequence information only. They use conservation measures (i.e. entropy) or evolutionary models (i.e. substitution matrix).

There is a list in this review (not OpenAccess, but ask me for a copy):

http://www.biochemj.org/bj/449/bj4490581.htm

Here is the list (some are using 3D structure information):

Tools URL

3DSim http://3dsim.bioinfo.cnio.es/

Badasp http://www.southampton.ac.uk/~re1u06/software/badasp/

CAPS http://bioinf.gen.tcd.ie/caps/

CHAIN http://www.chain.umaryland.edu/

CheckCov (available from request to the authors)

ConSurf http://consurf.tau.ac.il/

Diverge http://xungulab.com/software.html

Evolutionary Trace http://mammoth.bcm.tmc.edu/

FastML http://fastml.tau.ac.il/index.html

FunDi http://rogerlab.biochemistryandmolecularbiology.dal.ca/fundi.php

GroupSim http://compbio.cs.princeton.edu/specificity/

HMMDiverge http://info.mcmaster.ca/yifei/software/HMMDiverge.html

HyPhy http://www.hyphy.org

Meme http://www.hyphy.org/w/index.php/MEME

MultiRELIEF http://www.ibi.vu.nl/programs/multirelief/

Mutationassessor http://mutationassessor.org/

PAML (CodeML) http://abacus.gene.ucl.ac.uk/software/paml.html

Procov http://www.mathstat.dal.ca/~hcwang/Procov/

Rateshift http://www.daimi.au.dk/~compbio/rateshift/

Rate4Site http://www.tau.ac.il/~itaymay/cp/rate4site.html

RAte Shift EstimatoR (RASER) http://www.tau.ac.il/~penn/raser.html

SAAPdb http://www.bioinf.org.uk/saap/

SDPpred http://bioinf.fbb.msu.ru/SDPpred/

Selecton http://selecton.tau.ac.il/

Shift-Finder http://sites.univ-provence.fr/evol/softwares/shiftFinder/SHIFT-FINDER.htm

SIFT http://sift.jcvi.org/

SPEER http://www.hpppi.iicb.res.in/ss/

SPEL ftp://iole.swmed.edu/pub/SPEL/

TDG09 http://www.homepages.ucl.ac.uk/~ucbtaut/

TestNH http://biopp.univ-montp2.fr/forge/testnh

Xdet http://pdg.cnb.uam.es/pazos/Xdet/

Evolutionary Databases URL

Cube-DB http://epsf.bmad.bii.a-star.edu.sg/cube/db/html/home.html

FunShift http://funshift.sbc.su.se/

FunTree http://www.ebi.ac.uk/thornton-srv/databases/FunTree/

Human PAML browser http://mendel.gene.cwru.edu/adamslab/cgi-bin/paml/pbrowser.py

SDR http://paradox.harvard.edu/sdr

Selectome http://selectome.unil.ch/

The Adaptive Evolution Database (TAED) http://www.wyomingbioinformatics.org/TAED/

In a previous study (in OpenAccess), I used BADASP and CheckCov:

http://mbe.oxfordjournals.org/content/27/11/2618.short

I have also made a tutorial on how to use some of these tools:

http://www.cathdb.info/wiki/doku/?id=tutorials:eccb_t2_studer

Of course, all the tools are making prediction. So you cannot say for sure which amino acids are responsible for functional divergence or not. But that can help to select a few ones for mutagenesis for exemple.

Don't hesitate to contact if you need any help!

Best regards,

Romain

--

Romain Studer

EMBL-EBI

Wellcome Trust Genome Campus

Hinxton

Cambridgeshire

CB10 1SD, UK

Twitter: @RomainStuder

If you want to use BADASP and need any help with it, let me know. I also have a commandline program for making high(isn) quality input alignments and trees of subfamilies for it (HAQESAC), which can probably also be used for other such programs. See http://www.ncbi.nlm.nih.gov/pubmed/17220901 and http://bioware.soton.ac.uk/ if you are interested. The whole package and documentation is undergoing a bit of an overhaul at present but it's not too hard to get working with a little assistance! :op

PS. The link to BADASP that Romain posted is currently broken as it has moved but I will try to get a redirect up!

Dear Romain,

Thank you very much for the overwhelming list of options. I have difficulty seeing forest from through the trees. Nonetheless, as a newcommer in the field of functional divergence analysis I would certainly be unterested to learn more about it . Therefore, I would like to contact you personaly through my email: [email protected].

Dear Richard,

Another kind support in the field of bioinformatics. You mentioned BADASP as being useful for functional divergence, but why this program in contrast to all others?

BADASP has many advantages in my opinion:

- Provide both Type I and Type II of functional divergence.

- Use a phylogenetic tree to predict ancestral states.

- Easy to run.

- Made in python (so working in Windows, MacOX or Unix).

- Full command line, makes it easy to use it in pipeline.

One disadvantage is there is no statistical test a posteriori (i.e. likelihood value, baysian test). But we can run simulation to adjust the threshold for the predicted sites.

Romain

PS:

- Type I is when a site is conserved in a paralogous subgroup, but highly divergent in the other paralogous subgroup. This is also called heterotachy. covarion-like or rate-shifting sites (RSS).

- Type II is when a site is conserved in a paralog subgroup, and also conserved in the other paralog, but for another amino acid type or physico-chemical property. This is aslo called constant-but-different (CBD), conservation shifting sites (CSS) or heteropecilly.

Dear Rene,

Sorry for the delayed response - currently on the road. I would certainly defer to Romain's expertise here, as I have neither worked in this area for a while nor done a systematic review or comparison. BADASP was written to be automated and easy to understand. My opinion at the time was that there were a number of different methods available but none was clearly superior to another and reliable benchmarking datasets were rather thin on the ground.

If you have one family that you care a lot about, it is probably worth trying a few different methods that take different approaches. Even more importantly, take some time to make sure that you have a good sequence alignment, as this will be crucial to all methods. (Bear in mind that there might be some larger-scale differences responsible for functional differences, which could be missed by methods that purely considered conservation patterns of individual amino acid positions shared by all members. BADASP will compare amino acids to gaps but it is a bit of a fudge.)

Rich

PS. I've noticed a bunch of broken links on the website - I'm yet to catch up with two institution moves and a reorganisation of the code. Just DM me if you decide to use it and have any problems.