This study presents an integrated approach for sustainable food supply chain design by evaluating how sourcing geography and logistics network structure influence Global Warming Potential (GWP) in a multi-echelon Mexican cold chain integrating Life Cycle Assessment (LCA) and Linear Programming (LP) network optimization. Three soy products are evaluated: edamame from China, tofu from the U.S., and textured vegetable protein (TVP) modeled as a soy-based alternative. Results are calculated using a cradle-to-retailer system boundary, normalized to 100 g of delivered protein. Four network configurations are evaluated, varying sourcing geography, port selection, and warehouse allocation. Distribution-stage emissions are minimized through LP optimization, while upstream emissions are incorporated as exogenous LCA parameters. Sourcing geography, distribution-network design, and protein density significantly affect GWP per functional unit, with domestic sourcing yielding the lowest impacts for all products and network configurations. Tofu under the baseline configuration exhibits the highest GWP (1.2236 kg CO₂e/100 g protein), whereas TVP with domestic sourcing exhibits the lowest (0.1146 kg CO₂e/100 g protein), representing a 90.64% difference. The integrated approach provides a decision-support framework for lower-emission sourcing and distribution in emerging-economy food supply chains.