Integrated computational profiling of disorder propensity, adaptive evolution and structural confidence in classical swine fever virus proteins

Classical swine fever virus (CSFV) encodes a compact polyprotein whose mature proteins must retain essential structural functions while accommodating limited adaptive variation. Here, 212 complete CSFV polyprotein sequences were partitioned into 12 mature proteins and analysed using an integrated computational workflow combining IUPred3/ANCHOR2, metapredict, entropy profiling, IQ-TREE phylogenetics, HyPhy FEL/MEME/FUBAR selection tests, GARD recombination screening and ColabFold/AlphaFold2 pLDDT mapping. C/capsid and Npro showed the strongest support for classical disorder or disordered-binding propensity, whereas E2, NS5A and NS3 were better interpreted as structured or partially structured proteins carrying localized candidate adaptive-signal regions. Purifying selection dominated the proteome, but cross-method analysis prioritized 18 moderate/high-confidence candidate residues and three high-priority residues, with the strongest convergence in NS5A. GARD analysis of E2, NS5A, NS3 and NS5B did not indicate strong recombination-associated phylogenetic incongruence as the primary explanation for the main adaptive signals. Structural-context mapping placed the final candidate residues on rank-1 ColabFold models without treating them as experimentally validated functional sites. Overall, this study provides a reproducible computational framework for prioritizing CSFV protein regions for future antigenicity, immune-modulation and replication-complex studies.