L004: Cro66 and His148 bonding with regression¶
Methodology¶
Dihedral angles were transformed using the function,
This transformation maps the circular dihedral data to a [0, 1] range, preserving the periodicity while differentiating between cis (0°) and trans (180°) conformations. All input features (\(X\)) were standardized using sklearn's StandardScaler to ensure each feature contributes equally to the model. The distance between Cro66 OH and His148 HD1 (\(y\)) was used as the response variable without scaling.
Feature importance¶
Consistently low ranked: Tyr145 \(\phi\), Glu222 \(\phi\), Glu222 \(\psi\), Ser205 \(\psi\).
Regression on all states' trajectories.
Frequent disagreements.
Regression on only trajectories from reduced simulations.
Frequent disagreements; especially on Thr203 \(\psi\), Thr203 \(\phi\), and many others.
Regression on only trajectories from oxidized simulations.
Agreement on top one: Cys147 \(\phi\).
Disagreement on: Thr203 \(\psi\), Cys204 \(\phi\), Thr203 \(\psi\).
Regression on only trajectories from Cu(I) simulations.
Consensus on top two: Cys147 \(\psi\) and His148 \(\phi\).
Large differences in: Thr203 \(\psi\) and Thr203 \(\phi\).
