Fig. 10

Performance of MIOSTONE in transferring knowledge from pre-trained models in terms of AUROC. a A model on the large HMP2 dataset is pre-trained and then employed for the smaller IBD dataset in three settings: direct prediction on IBD (i.e., zero-shot), fine-tuning on IBD, and training IBD from scratch. Only tree-aware methods and MLP are included in the comparison, as most tree-agnostic methods are not well-suited for fine-tuning. Among the tree-aware methods, Ph-CNN is excluded because it is not scalable for processing the large HMP2 dataset. The prediction is conducted across three settings 20 times with varied train-test splits, and reported by the average performance assessed by AUROC, along with 95% confidence intervals. For scientific rigor, the performance between fine-tuning and training from scratch is quantified using one-tailed two-sample t-tests to calculate p-values. b The training dynamics of various models were evaluated by comparing fine-tuning with training from scratch, analyzing AUROC on test splits across different training epochs. MIOSTONE’s fine-tuning achieved better performance than training from scratch, requiring fewer training epochs