114 A Practical Guide to Cancer Systems Biology
on related technologies.^2 Fortunately, some researchers developed methods
to deduce protein structure computationally, trying to solve such problems.
Currently, methods of computational protein structure deduction can be
separated into two categories:ab initio(which is “from the beginning” in
Latin) and homology modeling. Methods classed asab initiomethods will
take protein as their sole input, model protein structure with knowledge of
physics and biochemistry from scratch by using techniques such as molecular
dynamics, and output the model. In the work of Chowdhuryet al.,^3 part
of miniprotein tc5b is folded by using molecular dynamics, and the result
highly resembles the structure from nuclear magnetic resonance (NMR).
However, as for now,ab initiomethods are not reliable, and cannot easily
work in large proteins since they will consume too much computational
resources.4,5In the second category, homology modeling, a protein model
is created from similar proteins with structures. That is, query sequence
is compared with other similar sequence(s) with structures. The homology
modeler will then assemble protein models according to the similarity. The
more identical residual exists between the template and the query, according
to their sequence alignment, the more accurate produced structure is.^6 In this
chapter, we focus on homology modeling (Fig. 1), since it’s the most common
technique used in protein modeling, and introduce some useful tools used in
homology modeling procedure.
In homology modeling, first, one should obtain protein sequence of the
target protein from a protein database, such as UniProt^7 or NCBI Protein.^8
Second, an appropriate template should be selected from structure database
like Protein Data Bank and alignment of query sequence and template
should be created. Then, the alignment will be submitted to the modeling
software such as SwissModel,^9 I-tasser,^10 or ModWeb.^11 In some modeling
software package, template search and modeling are automatically done
without user interference. However, before running homology modeling, some
model databases, for example, ModBase^12 and Protein Model Portal,^13 can
be searched for models created by other researchers. Finally, to check if
quality of the models is acceptable, protein quality assessment tools can be
applied on the models. ModEval^14 and Qmean server^15 can do protein quality
assessment for those who want to check their protein structure models.
- Checking existing computational protein models
Before you dive into homology modeling, checking if existing model exists is
recommended, since homology modeling is a time-consuming process. Such
checking may prevent unnecessary resource waste.