I want to find out the overall protein sequence similarity among 2 strains of same bacterial specie.
Let say, strain 1 has 4500 proteins and strain 2 has 4300. My strategy was; I performed all-against-all BLASTp of both the strains and selected the best-hit from strain 2 against the query from strain 1. Then sorted them according to their percent identity values. But. I am getting a range of NO HITS to 100% similarity. I was expecting to get a range of 80 to 100% since both the strains belong to same specie.
Kindly guide me to improve my current strategy or suggest me some computational tools to perform similar task.
AOA Suffian,
Try UniProt for aligning your twi sequences:
http://www.uniprot.org/align/
It has various options for annotation. May be that helps.
Regards,
FAZAL.
WAS Fazal bhai. Thank you for the suggestion.
I want to align two complete proteomes, not proteins. ALIGN will compare proteins. These are two different aspects.
What you see depends on the fact that you are not selecting for orthologous relationships, therefore in a certain number of cases you are recording the identity level among two paralogs. Since paralogs can pre-date the separation of the two strains under analysis, this decreases your average identity levels. The no hits depend on the query sequence (e.g. proteins shorter than 20 aminoacid can even fail to find themselves, depending on the size of the blast database) and clearly on the presence of a close-enough homologus sequence in the other proteome.
The best way to do it would be to run an orthologous classification algorithm (such as orthoMCL, Inparanoid, or simply a bidirectional best blast hit approach could suffice) and only then compare pairs of orthologs. Please, remember also that when you use blast, the alignments you get are local, and the identity percentage correspond to only the most significant HSP among two sequences. Therefore you should compare the pairs of proteins using a global alignment tool, like clustal, muscle, mafft or t-coffee and only then calculate the identity percentage.
Even in this case you will get a bell-shaped distribution whose shape will depend on the evolutionary distance among the two organisms. You are observing the effect of the different evolutionary rates specific of each protein or protein-family.
PS Then, if you really care about the matter, you also have to think that the percentage of identity is a quite raw measure of similarity because it does not depend on an evolutionary model (it only tells you how many identical residues there are in the alignment) therefore it weights in the same way every difference (both e.g. D->E and D->Y matter the same, which is clearly false from the biological point of view). You can some how alleviate this by using the similarity percentage (this depends from an evolutionary model telling which residues have to be considered similar) or going into the hardcore, by doing phylogenetic analysis and using the branch lengths expressed e.g. in substitutions per position.
One tool that was made for comparing the complete genomes of two isolates of the same species of bacteria, is Double-ACT. ACT and Double-ACT work on DNA genomes rather than on proteomes, but most proteome data is derived from DNA data anyway, so I am not sure I can think of a reason why using DNA would be detrimental. If ACT and Double-ACT do not work well on your pair of isolates, it may be wise to attempt to put the "species" you are working on in perspective of a few other common bacterial "species". The level of diversity that is considered to be different strains of the same species in one type of bacteria is far different than for other types. For example there are many bacteria classified into "groups" of Clostridium botulinum while the diversity within any one of those groups is greater than the diversity between Escheichia coli and Salmonella enteritidis. Modern molecular analyises of bacteria are proving that in some cases the classical classification methods could be quite misleading.
A second step is to consider which classes of genes/proteins are giving results near 100% amino acid identity and which are giving more distant results. With two isolates within the same species, we would expect the core metabolic function genes/proteins (ribosomal RNA and proteins, DNA and RNA polymerases, energy production enzymes, etc) to be very highly similar, while genes which are horizontally transferred more often (phage genes, transposases, helicases, integrases, etc) to be highly variable or present/absent between the two isolates.
In the case of the various groups in the Clostridium botulinum "species" the classification was based on the production of toxin but the toxin operon has been horizontally transferred between very different genomes.
http://www.hpa-bioinfotools.org.uk/pise/double_act.html
http://bioinformatics.oxfordjournals.org/content/21/16/3422.full
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools