Motivation: Protein networks are commonly used for understanding how proteins interact. However, they are typically biased by data availability, favoring well-studied proteins with more interactions. To uncover functions of understudied proteins, we must use data that are not affected by this literature bias, such as single-cell RNA-seq and proteomics. Due to data sparseness and redundancy, functional association analysis becomes complex. Results: To address this, we have developed FAVA (Functional Associations using Variational Autoencoders), which compresses high-dimensional data into a low-dimensional space. FAVA infers networks from high-dimensional omics data with much higher accuracy than existing methods, across a diverse collection of real as well as simulated datasets. FAVA can process large datasets with over 0.5 million conditions and has predicted 4210 interactions between 1039 understudied proteins. Our findings showcase FAVA's capability to offer novel perspectives on protein interactions. FAVA functions within the scverse ecosystem, employing AnnData as its input source.