In　this　paper,　we　consider　inference　aspects　of　skew-normal　semiparametric　varying　coefficient　models　which　provide　a　useful　extension　of　the　normal　regression　models.　The　maximum　likelihood　estimation　based　on　B-spline　is　proposed.　Further,　we　discuss　the　score　test　for　homogeneity　of　the　variance　in　skew-normal　semiparametric　varying　coefficient　models.　Their　asymptotical　properties　are　investigated.　Some　simulated　examples　are　used　to　examine　our　proposed　methods.