Fig. 3: Performance comparison between AdaMBind and baseline models under novel task split based on sequence similarity.
From: A meta learning and task adaptive approach for drug target affinity prediction
a, b Performance evaluation of AdaMBind and baseline models under the majority setting (a) and few-shot setting (b). In the majority setting, the support set size is 40, while in the few-shot setting, it is 5. The evaluation metrics include MSE, CI, R2, Spearman and Pearson. Five independent replications of each method were performed (n = 5). Data are expressed as means ± std. c Heatmap showing performance gains or losses of AdaMBind compared to its strongest competitor under both majority and few-shot settings. The three subplots correspond to 5 metrics (MSE, CI, R2, Spearman, Pearson). The y-axis lists different datasets, and the x-axis represents either the majority or the few-shot setting. Color represents performance difference, where red indicates performance improvement and blue indicates decline. Source data are provided as a Source Data file.
