2.5. Pipeline Alternatives

While matching results is an easy way to ensure pipeline quality, improvements may also result from comparing the current structure, originating from Mehmood et al., to those of the other pipelines.

Across pipelines, PSORTb is the most popular tool for localization, with some papers choosing to complement its results with CELLO. BlastP is usually used over DIAMOND, and VFDB is widely cited. In some works, VFDB was complemented with the microbial virulence database MvirdB (note: last updated snapshot in 2017) and the VirulentPred database (released in 2023).

The most variety can be found in the first part, the ‘homology filter’, with pipelines using a wide range of tools from the Mehmood et al. to in-house pipelines. Essential proteins are usually extracted using bioinformatics tools, most of which require licenses or are only able to handle limited amounts of data for short runtimes. used Clusters of Orthologous Groups (COG) distribution analysis, which, apart from how conserved proteins are, also shows their functional group (e.g., metabolism), allowing for better filtering in the first step (which comes at the expense of more data and run time required). Steps are otherwise similar, with some pipelines adding additional steps towards the end. Namely, evaluated antigenicity of many candidates in animal models, and performed physicochemical analysis of vaccine constructs and mapped B/T cell epitopes and their antigenicity/virulence prediction using ABCPred, ProPred1, and ProPred. For the K. pneumoniae pipeline, molecules were also screened using BepiPred-3.0, DiscoTope, and NetMHCII using specific information related to the mouse species, the lab would test the antigens.