tabix .

Just to elaborate a little bit Pierre's answer...

I agree that tabix would be the fastest way to do it, but considering that dbNSFP comes in a single zip file containing chromosome gzipped (not bgzipped) files, the best way to do it would be to unzip the particular chromosome you're interested in, bgzip it, tabix index it, and then query it with tabix.

Say you're interested in position 21:5011803, then you should go for something like this:

unzip -p dbNSFP4.1a.zip dbNSFP4.1a_variant.chr21.gz \
| gunzip | bgzip > dbNSFP4.1a_variant.chr21.gz
tabix -b2 -e2 -S1 dbNSFP4.1a_variant.chr21.gz
tabix dbNSFP4.1a_variant.chr21.gz 21:5011803-5011803

You could even go for a simple grep if you don't want to generate any intermediate files:

unzip -p dbNSFP4.1a.zip dbNSFP4.1a_variant.chr21.gz | zcat | head -1 > result.tab
unzip -p dbNSFP4.1a.zip dbNSFP4.1a_variant.chr21.gz | zgrep -P "^21\t5011803\t" >> result.tab