Blastp for identifying genes in contigs

[vanillasky]

I have used NCBI blastp as a first step to identify predicted ORFs in my assembled contigs. My question is how to filter these results based on % identity of the protein and e-value scores. I picked a loose protein cut-off of 20% but I'm not sure how one goes about selecting the best filtering options.
Any input would be appreciated. Thnx.

[maubp]

I would output the BLAST results in tabular format, and then filter using a simple script. Tabular output is quite easy to work with. You could even do this in Excel if you really prefer.

[vanillasky]

Cut-off values

Thank you for your suggestion. However, my question is a fundamental one. How does one decide on a cut-off value (s) with which to filter out the blasp results? For example based on greatest % identity for a protein match or an e-value score?

[rhinoceros]

Check the discussion here.

[maubp]

I don't have access to the paper right now (I'm away from the office), but I think reading this would be useful:

Punta and Ofran (2008) The Rough Guide to In Silico Function Prediction,
or How To Use Sequence and Structure Information To Predict Protein
Function. PLoS Comput Biol 4(10): e1000160.

http://dx.doi.org/10.1371/journal.pcbi.1000160

[GenoMax]

@vanillasky: Many things informaticians do can be automated/done in high throughput mode but once you get to "annotation" it may be best to slow down and do things the right way (that is why "finishing" a genome takes so long).

You could choose a value ("n") for the % cut-off and get a good approximation of the identities of proteins in your dataset. Afterwards, only a fraction (more or less, would depend on how good your blast results were) of those identities may turn out to be correct on more rigorous examination (as an example http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2673347/).

In general you can be reasonably sure about the validity of a protein blast if the e value is < 10^-3 and sequence identity is >= 25%.