Question

E-value for microbial computational genomics/genetics?

Particularly need help with 1a, but if you can help with any of the rest that would be great!

Thanks in advance!

1. As discussed in the final computational lecture, PSI-BLAST' is an iterative multiple sequence alignment technique that starts with a single protein query of database search by pairwise alignment. The resulting pairwise alignments from that search are collated and analyzed to create a PSSM (position specific substitution matrix) which is subsequently used in place of the original sequence to search the database. A PSSM is something like a consensus sequence that describes the protein family more completely than the original query alone (and therefore better at catching more distant members) but also risks targeting non-homologs if the PSSM becomes contaminated by unrelated sequences. Upon each iteration, new sequences may be found which are then optionally incorporated into the PSSM for the next search. The overall E-value from this iterative process should be estimated as follows: Each time the PSSM is updated, the highest E-value of any sequence included in the PSSM is added to the overall E-value of any final resulting alignment ("hit"). Thus, if a standard threshold for each iteration Ethresh is employed, then the overall E-value for a final alignment after Niter iterations can be expressed as: E = Niter x Einresh + Ehit A quick example: Suppose you run PSI-BLAST with a threshold of 0.005, meaning that no sequences with an E-value larger than 0.005 are included in each iteration. After six iterations, you find a promising "hit" (alignment) with E = 1 x 10-20. The overall E-value for that alignment would be estimated to be 6 x 0.005 + 1 x 10-20 = 0.03. a. For the example above, conceptually speaking, why is the overall E-value so large, even though the final alignment reported an E-value of 1x10-20? [2 points]

Answer 1

a) Overall E-value is directly proportional to the number of iterations. Thus, the overall E-value score actually is the score of the sequence that was obtained after 6 iterations with a significant E-value. The overall E-value of obtained after n iterations is always going to be higher than the (n-1)th iteration.

b) option c.

c) 0.1*2 + 1*10^-4 = 0.2 (Overall e-value score)

d) Low e-value score to obtain homologous sequences and reduce false positives.