Project Optimus, initiated by the US Food and Drug Administration (FDA), seeks to shift the focus of dose finding and selection from the maximum tolerated dose to the optimal dose that offers the most favorable risk-benefit balance. However, applying this paradigm shift to drug combination trials presents challenges, particularly due to limited sample sizes and a large two-dimensional dose exploration space. These challenges are amplified when trials involve multiple indications. To address this, we developed a two-stage Bayesian dose optimization design, called COMIC (Combination Optimization in Multiple IndiCations), to efficiently identify Optimal Biological Dose Combinations (OBDC) for multiple indications. The COMIC design follows a two-stage strategy: First, optimizing the dose for one indication based on a utility function that measures the risk-benefit tradeoff, and then using that data to inform and accelerate dose optimization for additional indications. This approach significantly reduces the required sample size. Additionally, we incorporate a pharmacodynamic endpoint (e.g., receptor occupancy) to prioritize which component of the combination should be escalated, further enhancing the efficiency of dose optimization. Simulation studies demonstrate the strong performance and robustness of the COMIC design across various scenarios. We illustrate the method using a CAR-T therapy trial.