As　a　spectral-efficient　and　transmission-reliable　technique,　channel　coded　layered　asymmetrically　clipped　optical　orthogonal　frequency-division　multiplexing　(LACO-OFDM)　has　attracted　considerable　interest　in　visible　light　communications.　In　this　letter,　we　investigate　the　protograph　low-density　parity-check　(LDPC)-coded　LACO-OFDM　VLC　systems.　Specially,　we　design　a　novel　type　of　interleaving　scheme,　referred　to　as　&　quot;variable-node　subcarrier　matched　mapping　(VNSMM)&　quot;　to　optimize　the　performance　of　protograph-coded　LACO-OFDM　VLC　systems.　In　such　scenario,　the　conventional　iterative　detection　framework　cannot　completely　eliminate　the　inter-layer　interference　(ILI),　which　degrades　the　system　performance.　To　further　suppress　the　ILI　at　the　receiver,　we　propose　a　deep　learning　(DL)-based　detection　framework,　called　&　quot;DeepLACO　&　quot;,　to　simultaneously　obtain　the　log-likelihood　ratios　(LLRs)　used　for　the　protograph　decoders　in　all　layers　without　the　knowledge　of　channel　state　information　(CSI).　Analyses　and　simulations　demonstrate　that　the　protograph-coded　LACO-OFDM　exploiting　the　proposed　interleaving　scheme　and　detection　framework　can　achieve　excellent　performance　in　VLC　systems.