Hello group,
I had predicted peptide sequences from denovo assembled contigs using abinitio (GENSCAN) approach and subjected it to similarity (BLASTP) search to identify genes in the assemble sequences. But the difficulty i am facing is with minimum percent of identity and coverage of blast hits. What should be the minimum threshold for percent identity and coverage so that it can be said for sure that the gene is present? This is a eukaryotic genome data.
Hey,
Since I've already done this (PMID: 26346253), I know what's your difficulty.
Firstly, did you blast all the sequences against NR database or organism specific database?
If its NR database, then considering coverage greater than 40% and identity greater than 95% would be better. I would even segregate them like query coverage x and identity y and then look for domain-wise identity to confirm on the query protein.
If it's against a specific organism's database, query coverage of >95% and identity >95% (cause the 5% can be considered as sequencing error) would be considered better.
Thanks for sharing your experience,
I had searched all predicted peptides against uniprot database to identify genes and its function. I came across paper stating, E-value (~10-6) values and bit scores (bits > 50) are far more sensitive and reliable than percent identity for inferring homology.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3820096/
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools