ZWINT Interactome Exploration: Supporting Data for a Novel AI-Integrated PPI Workflow

<b>Introduction: </b>ZWINT (ZW10 Interactor) is known conventionally for its part in kinetochore movement during mitosis; however, it has also been implicated in neuropathic pain pathways, making its interactome a growing region of interest. Given the limited number of known Zwint PPIs within neurons, a substantial amount of real-world validation would need to be performed. Our goal with this study was to develop a workflow using various in-silico tools, including the AI protein folding "AlphaFold", that would allow for the development and prioritization of PPIs through computational triage. The associated short communication describes this novel methodology, with this data serving as proof-of-concept evidence. <b>Selected Proteins: </b>A total of seven proteins were tested to see how well they might bind to ZWINT. Three of these were known interactors determined through a mass spectrometry experiment, as characterized on ZWINT's Biogrid page. These interactors served as positive controls and were as follows: SNAP25CAMK2AUBCThe four remaining proteins were selected due to their proximity to ZWINT, including having known interactions with proteins already interacting with ZWINT. VCP, interacts with UBCARC, interacts with CAMK2ASTX1A, interacts with SNAP25BLOC1S2, implicated in the same neuropathic pain study as ZWINT <br><b>Methodology:</b>This dataset supports the development of a two-pathway workflow to prioritize specific PPIs. The approach combines a precision arm where dimers are created by AlphaFold and RoseTTAFold before their similarity is assessed by superimposing them in TM-Align. The AlphaFold structure with the highest similarity to the RoseTTAfold structure is then used for the binding arm of the study, where the interface of the protein complex is determined via a specialized "interface residues" command in the software "Open Source PyMOL". Following this, the structure and residue numbers are input into HADDOCK to get a HADDOCK score, where a lower score correlates to higher binding potential. <b>Disclaimer: </b>This process is not intended to use AI to replace real-world research. Rather, it is meant to triage already existing PPI candidates or give ideas for preliminary results that can be used in hypothesis generation before confirming with real-world experimentation. Note that the exploratory candidates here have NOT yet been experimented with in a real-world setting, and instead were created to see how their TM scores and HADDOCK scores fared compared to the positive controls.