Gram-negative bacteria have evolved an extraordinary array of secretion systems to export protein substrates into target cells or the surrounding environment. These protein substrates differ significantly in their structures and functions, and in the dedicated secretory pathway that they use. Accordingly, it is difficult to develop computational systems for accurate prediction of substrate type.

In this work, we present an integrative system, BastionX, to predict and visualize various types of secreted substrates in Gram-negative bacteria. BastionX outperforms existing substrate predictors by four major upgrades: 1) BastionX is the first predictor of types 1 and 2 secreted substrates, and also includes more accurate predictions for types 3, 4, and 6 secretion systems; 2) integrating single predictors within a framework, BastionX can seamlessly annotate various types of secreted substrates in a given bacterial genome and provides interactive visualization for its general substrate distribution; 3) Specifically, for each predicted substrate, BastionX enables further functional analysis within four interactive visualizations; and 4) BastionX provides three modes to cater to different needs: Fast mode for rapid prediction, Accurate mode for accurate prediction, and Balanced mode for a balance between prediction accuracy and speed. In combination with its additional standalone toolkit, BastionX can be conveniently executed locally to conduct sequence analysis and be readily integrated into a user’s own pipeline. Taken together, BastionX can simultaneously annotate thousands of protein sequences with their potential secretion substrate type and map the global landscape of how secreted substrates are distributed across bacterial genomes.