SBGNFlow: An AI-Assisted & Interactive Workflow for Generation, Merging & Layout of Pathway Maps - Quantitative Testing Dataset

This dataset contains the resources used in the quantitative evaluation of the sketch-based SBGN generation component of SBGNFlow, available at https://github.com/sciluna/sbgn-flow. It includes 60 hand-drawn Systems Biology Graphical Notation (SBGN) diagrams (30 Process Description (PD) and 30 Activity Flow (AF)) together with the in-context learning (ICL) materials used in the evaluation. For each evaluation diagram, the dataset provides: the hand-drawn sketch, the original digitally generated SBGN diagram, the corresponding ground-truth SBGN-ML representation, and a metadata file. The metadata files report the SBGN language, graph size, node and edge class distributions, and the associated filenames. The directory structure is organized as follows: PD/ handdrawn/ original/ sbgnml/ metadata/ AF/ handdrawn/ original/ sbgnml/ metadata/ ICL/ PD/ reference_card.png sample_handdrawn.png sample.sbgn AF/ reference_card.png sample_handdrawn.png sample.sbgn The hand-drawn sketch, original diagram, SBGN-ML file, and metadata file for each evaluation example share the same identifier. The ICL directory contains the reference cards and example diagram/SBGN-ML pairs used in the in-context learning experiments described in the SBGNFlow application note. These files are provided to facilitate reproducibility and are not part of the quantitative evaluation dataset. This dataset accompanies the SBGNFlow application note and is intended for benchmarking methods that convert hand-drawn pathway sketches into structured SBGN representations.