methods are successful in predicting likely mutations leading to drug resistance.
These methods are able to predict due to their amenability to decompose the free
energy into its components at the residue level that leads to better understanding of
the effect of mutations on drug binding. Lethal effects of the V82F/I84V double
mutation in HIV-1 protease on amprenavir were demonstrated using MM-PBSA
approach on snapshots obtained from the well-equilibrated protein–ligand complex
[102]. It was reported that amprenavir lost its binding affinity due to distortions in
the binding site, hence weakening many favourable interactions (DDG = 3.73 kcal/
mol). Such a distortion of the binding site was previously observed and attributed to
the rapid flap movements seen in this double mutant which is absent in the
wild-type HIV-1 protease [103]. Furthermore, newer inhibitors, that are very close
structural analogues of amprenavir, like TMC126 (DDG = 2.01 kcal/mol) and
TMC114 (darunavir, DDG = 3.45 kcal/mol) were also seen to be affected by these
mutations, though to a lesser extent than amprenavir. Despite structural distortions
in the binding site, it had no effect on the substrate binding, and hence, the catalytic
process was unhindered.
Hou et al. [104] combined MM-GBSA with the positional variability approach,
to modify Kollman’s FV value [105] to give a new scoring function also called FV
(Free energy/Variability) score. Using the FV score, they evaluated the binding of
six substrates that are hydrolysed by HIV-1 protease and confirmed Kollman’s
[105] observation that drug-resistant mutations are more likely to occur at less
conserved regions. The FV score reported by Hou et al. comprises two components,
one that reflects the binding energetics at the per-residue level, obtained by
MM-GBSA, and the second component is the sequence variability that represents
the conservation of amino acids at each position. Using this score, one can identify
amino acid residues that are crucial for substrate and inhibitor binding, and thus
classify the residues that are exclusively involved in substrate binding and those
that are exclusive for inhibitor binding. Such a classification when coupled with the
positional variability of amino acid residues can extract those positions with low
conservation and exclusivity for inhibitor binding; such positions are highly
amenable to mutations leading to drug resistance. Employing this method Hou et al.
confirmed their previous observation [102] that the V82F/I84V double mutations
are lethal for many FDA approved HIV-1 protease inhibitors, whereas TMC126 is
still active against this mutant.
3.3 Vitality Analysis
One of the primary drawbacks of the aforementioned methods to predict
drug-resistant mutations is their inability to accurately estimate the binding affinity
for the substrate molecule(s). The fitness cost of the mutation can be estimated by
gauging the change in the binding affinity of the substrate to its enzyme target; any
perturbation in the substrate binding is likely to affect the function of the enzyme.
Therefore, computing the catalytic efficiency of the enzyme before and after
Free Energy-Based Methods to Understand Drug Resistance Mutations
17
These methods are able to predict due to their amenability to decompose the free
energy into its components at the residue level that leads to better understanding of
the effect of mutations on drug binding. Lethal effects of the V82F/I84V double
mutation in HIV-1 protease on amprenavir were demonstrated using MM-PBSA
approach on snapshots obtained from the well-equilibrated protein–ligand complex
[102]. It was reported that amprenavir lost its binding affinity due to distortions in
the binding site, hence weakening many favourable interactions (DDG = 3.73 kcal/
mol). Such a distortion of the binding site was previously observed and attributed to
the rapid flap movements seen in this double mutant which is absent in the
wild-type HIV-1 protease [103]. Furthermore, newer inhibitors, that are very close
structural analogues of amprenavir, like TMC126 (DDG = 2.01 kcal/mol) and
TMC114 (darunavir, DDG = 3.45 kcal/mol) were also seen to be affected by these
mutations, though to a lesser extent than amprenavir. Despite structural distortions
in the binding site, it had no effect on the substrate binding, and hence, the catalytic
process was unhindered.
Hou et al. [104] combined MM-GBSA with the positional variability approach,
to modify Kollman’s FV value [105] to give a new scoring function also called FV
(Free energy/Variability) score. Using the FV score, they evaluated the binding of
six substrates that are hydrolysed by HIV-1 protease and confirmed Kollman’s
[105] observation that drug-resistant mutations are more likely to occur at less
conserved regions. The FV score reported by Hou et al. comprises two components,
one that reflects the binding energetics at the per-residue level, obtained by
MM-GBSA, and the second component is the sequence variability that represents
the conservation of amino acids at each position. Using this score, one can identify
amino acid residues that are crucial for substrate and inhibitor binding, and thus
classify the residues that are exclusively involved in substrate binding and those
that are exclusive for inhibitor binding. Such a classification when coupled with the
positional variability of amino acid residues can extract those positions with low
conservation and exclusivity for inhibitor binding; such positions are highly
amenable to mutations leading to drug resistance. Employing this method Hou et al.
confirmed their previous observation [102] that the V82F/I84V double mutations
are lethal for many FDA approved HIV-1 protease inhibitors, whereas TMC126 is
still active against this mutant.
3.3 Vitality Analysis
One of the primary drawbacks of the aforementioned methods to predict
drug-resistant mutations is their inability to accurately estimate the binding affinity
for the substrate molecule(s). The fitness cost of the mutation can be estimated by
gauging the change in the binding affinity of the substrate to its enzyme target; any
perturbation in the substrate binding is likely to affect the function of the enzyme.
Therefore, computing the catalytic efficiency of the enzyme before and after
Free Energy-Based Methods to Understand Drug Resistance Mutations
17