Mass spectrometry-based proteomics provides detailed molecular information about biological samples but typically measures mixtures of multiple cell populations. Computational cell-type deconvolution aims to estimate the relative contributions of different cell types from such bulk measurements, enabling more detailed interpretation of complex biological tissues. This project focuses on the development, optimization, and evaluation of computational methods for cell-type deconvolution using proteomics data. Building on previous work, including the proteoDeconv software framework, we aim to improve the accuracy, robustness, and applicability of proteomics-based deconvolution approaches. Particular emphasis will be placed on addressing challenges associated with proteomics data, such as missing values, technical variability, and differences between reference profiles and biological samples. The project involves computational benchmarking, simulations, and the analysis of publicly available and experimental proteomics datasets. By systematically evaluating methodological choices and developing improved computational approaches, we aim to facilitate more reliable estimation of cellular composition from bulk proteomics measurements. The resulting methods and workflows will support applications in immunology, disease research, and other areas where understanding cellular heterogeneity is important. The project is related to parts of proposal NAISS 2026/4-231, but focused on cell-type deconvolution, and not a direct continuation.