Swarm　intelligence　algorithms　have　been　successfully　applied　to　the　detection　of　functional　modules　in　PPI　networks.　As　the　increasing　of　the　PPI　network　size,　those　algorithms　will　cost　more　time　in　functional　module　detection.　In　this　paper,　we　present　a　novel　algorithm,　ACC-MLF,　which　combines　ant　colony　clustering　with　multilevel　framework　to　reduce　the　runtime　in　the　large-scale　PPI　networks.　First,　use　a　new　matching　strategy　to　coarsen　the　original　large-scale　PPI　network,　and　get　a　smaller　PPI　network.　Then,　use　the　ant　colony　clustering　algorithm　to　cluster　the　obtained　network.　Finally,　get　the　clustering　result　of　original　network　through　de-coarsening　and　use　the　refinement　to　avoid　the　result　from　falling　into　the　local　optimal.　Experiments　in　some　large-scale　networks　show　that　the　detecting　speed　of　ACC-MLF　has　significantly　improved　in　contrast　to　ACC-FMD,　and　ACC-MLF　can　get　better　clustering　results　in　some　evaluation　metrics　while　compared　with　ACC-FMD,　MCODE,　MINE　and　Core　algorithms.