This study was designed to systematically identify novel umami peptides in lager beer, clarify their molecular interactions with the T1R1/T1R3 receptor, and determine their specific effects on multidimensional sensory attributes. The peptides were characterized by LC-MS/MS combined with de novo sequencing, and 906 valid sequences were obtained. Machine-learning models (UMPred-FRL, Tastepeptides-Meta, and Umami-MRNN) predicted 76 potential umami peptides. These candidates were docked to T1R1/T1R3 with the CDOCKER protocol, producing 57 successful complexes. Six representative peptides-KSTEL, DELIK, DIGISSK, IEKYSGA, DEVR, and PVPL-were selected for 100 ns molecular-dynamics simulations and MM/GBSA binding-energy calculations. All six peptides stably occupied the narrow cleft at the T1R1/T1R3 interface. Their binding free energies ranked as DEVR (-44.09 ± 5.47 kcal mol(-1)) < KSTEL (-43.21 ± 3.45) < IEKYSGA (-39.60 ± 4.37) ≈ PVPL (-39.53 ± 2.52) < DELIK (-36.14 ± 3.11) < DIGISSK (-26.45 ± 4.52). Corresponding taste thresholds were 0.121, 0.217, 0.326, 0.406, 0.589, and 0.696 mmol L(-1) (DEVR < KSTEL < IEKYSGA < DELIK < PVPL < DIGISSK). TDA-based sensory validation with single-factor additions showed that KSTEL, DELIK, DEVR, and PVPL increased umami scores by ≈21%, ≈22%, ≈17%, and ≈11%, respectively, while DIGISSK and IEKYSGA produced marginal changes (≤2%). The short-chain peptides thus bound with high affinity to T1R1/T1R3 and improved core taste and mouthfeel but tended to amplify certain off-flavors, and the long-chain peptides caused detrimental impacts. Future formulation optimization should balance flavor enhancement and off-flavor suppression, providing a theoretical basis for targeted brewing of umami-oriented lager beer.