This is my first time trying to do molecular replacement to solve a protein structure. I am using the NCBI blastp program to find suitable search models. When choosing a search model, I understand that sequence identity is the most important -- but should I also take into account e-values or query coverage (i.e, what % of my query sequence is in the alignment) or any other factors for the purpose of molecular replacement? Thank you for your help!


Identity, Similarity, E-value are the key numbers, for the prediction of the 3D structure of proteins. The Length of the query and selected results should also be close to each other.

E-value: The first value to look at is indeed the E-value. For the most accurate predictions, E-values less than $10^{-100}$ are of extreme importance While E-values ranging between $10^{-50}$ to $10^{-100}$ come as the second priority. E-values less than $10^{-20}$ or so, should not be trusted. (As you know the larger amount of E-value accounts for random hits which may not have any evolutionary background with your query)

Similarity and Identity: The ranges for these amounts differ for each survey and should be assessed with other statistics. Needless to say, Identities below 5% for amino acid blasts and below 25% for nucleotide blasts are useless. For proteins identity above 25% is acceptable and 35-50% (and more) are of greater importance. Similarity plays a more important role in the overall shape and structure of the protein. "Total score" in Blast NCBI should suffice for these evaluations.

Length: The sequence length of your results should be reasonably near to one another. For example, having a sequence of 600 amino acids as your query, the acceptable range for the selected results (to eventually predict 3D structure with) should be 600 $\pm$ 30. Five percent difference is usually enough. Deviation from this would result in extra 3rd and 4th structures induced by other subunits.

Other considerations: In the case that your organism is known, pick the sequences which share a common evolutionary history with your organism of interest See Tree of life. The number of subunits and Symmetry types might be useful in some cases. Avoid selecting ab-initio based 3D structures in your results all the time and focus mainly on X-ray diffraction-based structures. If you're using the PDB website for data retrieval, Refinement Resolution is also extremely important. Try both CIF and PDB formats as input, as the results may sometimes improve. At last but not least, try blasting using curated databases. In blastp, refseq_protein , swissprot , pdb , refseq_select are ideal.

This is just a rule of thumb and should not be generalized as the exact numbers differ substantially from one query to another.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.