In continuation to my earlier thread (subject: sequence independent model building possible?), I have built model into density without knowledge of sequence for a data diffracted to 1.9A resolution.  Current R/Rfree is 15/19, phaser error=16.88 degrees with no Ramachandran outliers.

Is there a way we can differentiate between Glu/Gln, Asp/Asn and sometimes Thr/Val directly from density?  I have considered the local environment (hydrophobic/hydrophilic/polarity pockets, possible hydrogen bonds/other interactions, buried/exposed, etc...) in choosing one over the other confusing pairs of amino acids.  However, I am not absolutely certain in many places.

A BLAST of this sequence against all non-redundant protein sequence database yield highest hit of 80% sequence identity.  Hence, we are still not sure of sequence of the contaminant protein which got crystallised and want to decipher sequence directly from the structure.

Thanks for any pointers/suggestions,

Regards,

Kaushik

--