Understanding protein-protein interaction (PPI) network dynamics is key to understanding nominal and perturbed cell states. Here, we develop a new machine learning framework called Tapioca that allows for the study of PPIs in dynamics contexts at a global scale in ex/in vivo conditions. Furthermore, we optimized the thermal denaturation and lysis conditions used in the thermal proximity coaggregation (TPCA) methodology, one of the types of data Tapioca can use to make predictions. Using this optimized protocol and Tapioca, we investigated the temporal dynamics of reactivation from latency of the oncogenic gammaherpesvirus Kaposi’s sarcoma-associated herpesvirus (KSHV), identifying the host protein NUCKS1 as a factor promoting KSHV genome replication during lytic infection. Integrating this dataset with published TPCA datasets from the alphaherpesvirus herpes simplex virus type-1 (HSV-1) and the betaherpesvirus human cytomegalovirus (HCMV), we determined NUCKS1 to have a proviral role across all three herpesvirus subfamilies.

Signature peptides unique to human NUCKS1, as well as KSHV proteins RIR1 and RIR2, were manually curated and subsequently quantified by scheduled PRM analysis on a Thermo Q Exactive HF. Peptide spectral libraries were built with Proteome Discoverer 2.4, and PRM data was analyzed in Skyline.

PRM analysis of immunoaffinity purification (IP) of endogenous NUCKS1 protein, with rabbit IgG antibody as a control, both in biological triplicate. IP was performed 48 hours post-reactivation (hpr) of KSHV from latency in the iSLK.219 cell type. Files contain quantified peptides for human NUCKS1, as well as KSHV proteins RIR1 and RIR2. Data for library creation can be found on PRIDE: PXD041152.