Extended Data Fig. 1: Feature extraction for ML training.
a, Twenty-two 3D spatial conformation features extracted from molecular density maps based on geometric properties (for example, angle, width and length). b, Six foundational 2D chemical structure features. c, Pearson correlation coefficient (PCC) matrix analysis of the 28 features. d, Feature importance analysis of the twenty-eight features, revealing their relative contribution to the model’s predictive performance. e, Prediction formula of the best-performing ML model (Model 1): ypre = c1D1 + c2D2 + c3D3 + c4D4 + c5D5 + c0, with coefficients and feature descriptions listed in the table.
