Fig. 3. Performance of ML model for prediction of mean square radius of gyration, 〈R_g²〉, for class I and class II polymers.

In (A), an ML model is trained on 80% of the sequences, and the predictions for 〈R_g²〉 versus simulated 〈R_g²〉 are shown for the 308 held-out sequences; the r² is 0.953, and the MAE is 111.32σ². In (B), a model trained on class I polymers is applied to make predictions on the class II polymers; the r² is 0.895, and the MAE is 130.34σ². The coloring of the markers reflects the polymer composition as described in the “Data coloring by composition” section in Materials and Methods.