Fig. 4

Performance of eleven DANN modules in different subsets. aPerformance of original trained DANN modules in different subsets. b Performance of DANN modules re-trained in merged subsets. c Performance of re-trained DANN modules in original subsets. X-axis shows the prediction accuracy, while y-axis shows the sensitivity. Each symbol stands for the performance of a specific subset predicted by specific DANN module. Larger symbols are performance of eleven original DANN modules trained in their corresponding subsets, while small symbols show performance of original DANN modules in non-corresponding subsets or performance of re-trained DANN modules in various subsets. The subsets are represented by the colours filled in the symbols as follows: D4:black, D3–1:red, D3–2:green, D3–3:yellow, D3–4:blue, D2–1:pink, D2–2:cyan, D2–3:grey, D2–4:dark red, D2–5:dark green, D2–6:dar yellow. DANN modules are denoted by the shape of symbols as follows (All DANN-3x have dark edges, while all DANN-2x have red edges): DANN4:star, DANN-31:up triangle, DANN-32:down triangle, DANN-33:square, DANN-34:diamond, DANN-21:up triangle, DANN-22:down triangle, DANN-23:square, DANN-24:diamond, DANN-25:hex, DANN-26:circle