Good Example Sequence For Psi-Blast

Question

[Simon Cockell]

I'm trying to put together a relatively simple BLAST exercise for lab biologists as part of a bioinformatics workshop.

As part of this exercise, we want to demonstrate the power of PSI-BLAST over a simple non-iterative BLAST search. The NCBI provides a PSI-BLAST tutorial with good example sequences, but this was written in 2007, and they no longer provide the eloquent demonstration they once did(!)

So my question is: can anyone provide me with a suitable test sequence for PSI-BLAST, one that essentially "fails" in BLAST, but will return a promising hit within a few iterations of PSI-BLAST?

Supplementary question: Are we at a point where the data is becoming so redundant that PSI-BLAST is no longer an effective strategy?

[Jan Kosinski]

Some say that PSI-BLAST is deprecated and you should use hhblits instead: faster, much more sensitive, much more accurate, http://www.nature.com/nmeth/journal/vaop/ncurrent/full/nmeth.1818.html

[Larry_Parnell]

Simon, you should post these questions to Deanna Church at NCBI. I'm sure she'd have some informed thoughts. She's on Twitter, and I think she once used BioStar.

Answer 1

[Asaf]

We ran into exactly the same problem in our course. We used P09464 and P18902. We took it from http://margalit.huji.ac.il/ ->Resources -> SSSD (number 2). The table contains pairs of PDB entries with comparable structure that can be found using PSI-BLAST but not BLAST.

Answer 2

[Neilfws]

To answer your second question: yes, there is now far more sequence data available than there was when PSI-BLAST was first introduced. This means that a non-profile-based similarity search has a far higher chance of finding a hit. Whether the hits have any useful functional annotation is a whole other issue.

Many more profile-based searching methods have been introduced since the publication of PSI-BLAST too, such as those using HMMs.

That's not to say that PSI-BLAST is no longer useful; it's often used, for example, to find templates for protein structure homology modelling. Here's an example from the MODELLER website, using nsp16 protein from the SARS coronavirus.