albanese_dissertation.lof

\addvspace {10\p@ }
\addvspace {10\p@ }
\contentsline {figure}{\numberline {2.1}{\ignorespaces Free energy calculations can accelerate selectivity optimization.}}{21}{figure.caption.3}
\contentsline {figure}{\numberline {2.2}{\ignorespaces A CDK2/CDK9 selectivity dataset.}}{24}{figure.caption.4}
\contentsline {figure}{\numberline {2.3}{\ignorespaces A CDK2/ERK2 selectivity set}}{27}{figure.caption.5}
\contentsline {figure}{\numberline {2.4}{\ignorespaces Selectivity predictions suggest correlation in forcefield error.}}{30}{figure.caption.6}
\contentsline {figure}{\numberline {2.5}{\ignorespaces Correlation in force field errors between targets can significantly accelerate selectivity optimization.}}{33}{figure.caption.7}
\contentsline {figure}{\numberline {2.6}{\ignorespaces Reducing statistical uncertainty when force field error correlation is high improves optimization speedups}}{36}{figure.caption.8}
\addvspace {10\p@ }
\contentsline {figure}{\numberline {3.1}{\ignorespaces Relative alchemical free-energy calculations can be used to predict affinity changes of FDA-approved selective kinase inhibitors arising from clinically-identified mutations in their targets of therapy.}}{60}{figure.caption.15}
\contentsline {figure}{\numberline {3.2}{\ignorespaces Cross-comparison of the experimentally measured effects that mutations in Abl kinase have on ligand binding, performed by different labs.}}{67}{figure.caption.17}
\contentsline {figure}{\numberline {3.3}{\ignorespaces Comparison of 31 mutations for which phosphorylated and non-phosphorylated $\Delta K_{d}$s were available.}}{68}{figure.caption.18}
\contentsline {figure}{\numberline {3.4}{\ignorespaces Comparison of experimentally-measured binding free-energy changes ($\Delta \Delta $G) for 131 clinically observed mutations and 6 selective kinase inhibitors for which co-crystal structures of wild-type kinase with inhibitor are available.}}{74}{figure.caption.19}
\contentsline {figure}{\numberline {3.4}{\ignorespaces Figure caption}}{75}{figure.caption.20}
\contentsline {figure}{\numberline {3.5}{\ignorespaces TKI-by-TKI truth tables with increasingly large classification cutoffs.}}{78}{figure.caption.21}
\contentsline {figure}{\numberline {3.6}{\ignorespaces Physical modeling accuracy in computing the impact of clinical Abl mutations on selective inhibitor binding.}}{81}{figure.caption.22}
\contentsline {figure}{\numberline {3.6}{\ignorespaces Figure caption}}{82}{figure.caption.23}
\contentsline {figure}{\numberline {3.7}{\ignorespaces Predicting resistance mutations using FEP+ for inhibitors for which co-crystal structures with wild-type kinase are not available.}}{87}{figure.caption.24}
\addvspace {10\p@ }
\contentsline {figure}{\numberline {4.1}{\ignorespaces Abl kinase domain construct expression screen illustrates high sensitivity to construct boundaries.}}{121}{figure.caption.40}
\contentsline {figure}{\numberline {4.2}{\ignorespaces Expression yields of Abl kinase domain constructs for all constructs with detectable expression.}}{122}{figure.caption.41}
\contentsline {figure}{\numberline {4.3}{\ignorespaces Kinome wide search for expressible kinases.}}{125}{figure.caption.42}
\contentsline {figure}{\numberline {4.4}{\ignorespaces Fluorescence-based thermostability assay demonstrates many high-expressing kinases are well-folded.}}{130}{figure.caption.43}
\contentsline {figure}{\numberline {4.5}{\ignorespaces Fluorescence emission spectra as a function of the fluorescent ATP-competitive kinase inhibitor bosutinib demonstrates the presence of a well-formed ATP binding pocket.}}{134}{figure.caption.44}
\contentsline {figure}{\numberline {4.6}{\ignorespaces Expression yields for engineered clinically-derived Src and Abl missense mutants}}{140}{figure.caption.45}
\addvspace {10\p@ }
\addvspace {10\p@ }
\contentsline {figure}{\numberline {A.1}{\ignorespaces CDK2 adopts an inactive conformation in the crystal structure used for the CDK2/ERK2 calculations}}{162}{figure.caption.52}
\contentsline {figure}{\numberline {A.2}{\ignorespaces Correlation coefficient $\rho $ controls the shape of the joint marginal distribution of errors}}{163}{figure.caption.53}
\contentsline {figure}{\numberline {A.3}{\ignorespaces Correlation reduces the expected error for selectivity predictions}}{164}{figure.caption.54}
\contentsline {figure}{\numberline {A.4}{\ignorespaces Each replicate of the CDK2/CDK9 calculations yields a consistent RMSE and MUE}}{165}{figure.caption.55}
\contentsline {figure}{\numberline {A.5}{\ignorespaces Each replicate of the CDK2/CDK9 calculations yields consistent errors and correlation coefficient}}{166}{figure.caption.56}
\contentsline {figure}{\numberline {A.6}{\ignorespaces Each replicate of the CDK2/ERK2 calculations yields a consistent RMSE and MUE}}{167}{figure.caption.57}
\contentsline {figure}{\numberline {A.7}{\ignorespaces Each replicate of the CDK2/ERK2 calculations yields consistent errors and correlation coefficient}}{168}{figure.caption.58}
\contentsline {figure}{\numberline {A.8}{\ignorespaces The pooled replicates show good agreement in predictions for individual ligands}}{169}{figure.caption.59}
\contentsline {figure}{\numberline {A.9}{\ignorespaces The standard deviation for each edge is smaller than the estimated cycle closure uncertainties}}{170}{figure.caption.60}