There are challenges in the genetic diagnosis of rare diseases, and pursuing an optimal strategy to identify the cause of the disease is one of the main objectives of any clinical genomics unit. A range of techniques are currently used to characterize the genomic variability within the human genome to detect causative variants of specific disorders. With the introduction of next-generation sequencing (NGS) in the clinical setting, geneticists can study single-nucleotide variants (SNVs) throughout the entire exome/genome. In turn, the number of variants to be evaluated per patient has increased significantly, and more information has to be processed and analyzed to determine a proper diagnosis. Roughly 50% of patients with a Mendelian genetic disorder are diagnosed using NGS, but a fair number of patients still suffer a diagnostic odyssey. Due to the inherent diversity of the human population, as more exomes or genomes are sequenced, variants of uncertain significance (VUSs) will increase exponentially. Thus, assigning relevance to a VUS (non-synonymous as well as synonymous) in an undiagnosed patient becomes crucial to assess the proper diagnosis. Multiple algorithms have been used to predict how a specific mutation might affect the protein’s function, but they are far from accurate enough to be conclusive. In this work, we highlight the difficulties of genomic variability determined by NGS that have arisen in diagnosing rare genetic diseases, and how molecular modelling has to be a key component to elucidate the relevance of a specific mutation in the protein’s loss of function or malfunction. We suggest that the creation of a multi-omics data model should improve the classification of pathogenicity for a significant amount of the detected genomic variability. Moreover, we argue how it should be incorporated systematically in the process of variant evaluation to be useful in the clinical setting and the diagnostic pipeline.