Germline mutation is the ultimate source of evolutionary change and disease-causing variants. Understanding the rates and patterns of human mutation can help us learn about their molecular origins, uncover our evolutionary history and improve our ability to identify the genetic causes of human disease. With the advent of exome and genome data sets of parent-offspring trios there is an unprecedented opportunity to characterise mutations at an individual level and to harness the increasing sample sizes to identify disease-causing mutations. The goal of this thesis is to understand sources of variation in germline mutation and the contribution of these mutations to rare developmental disorders. These sources of variation encompass types of mutations that have been previously underrepresented in genetic research as well as individual mutation rates and spectra across individuals and parental origin. These analyses fall into three distinct projects. My first project in this dissertation focuses on the mutational origins and pathogenic impact of multi-nucleotide variants (MNVs). These are variants that fall within 20 base pairs of each other and are frequently misannotated in variant-calling pipelines. Using data from the Deciphering Developmental Disorders (DDD) study, I explore the pathogenicity of this type of variant and found that MNVs in protein-coding sequences can be more pathogenic than a single nucleotide variant even when the MNV falls within a single codon. I also estimate the MNV mutation rate, explore the mutational spectra of these variants and describe the contribution of de novo MNVs to severe developmental disorders. The next project focuses on identifying and characterising germline hypermutators. Using sequencing data from the DDD and 100,000 Genomes Project datasets across ~20,000 parent-offspring trios, I identified fifteen children with an unusually large number of de novo mutations. Eight of these appear to be due to a paternal hypermutator. I describe analyses to try and identify a genetic cause for this hypermutation. For two of the individuals, I found rare homozygous paternal variants that fell into two different DNA repair genes and are the likely cause. I also explore whether variants in DNA repair genes more generally impact germline mutation rates. First by examining a well characterised cancer somatic mutator gene and second by using a broader approach across all DNA repair genes. Using the large resource of DNMs called in the 100,0000 Genomes Project dataset, I also estimate what fraction of variance in germline mutation rate can be explained by hypermutation as well as by parental age. In my final project, I describe analyses of de novo mutations in a cohort of individuals with developmental disorders (DDs). De novo mutations are a major cause of DDs however known genes only account for a minority of the observed excess of these mutations. Here I develop a statistical framework and apply this on de novo mutations from ~31,000 exome sequenced parent offspring trios from the DDD study pooled with trios from GeneDx, a US-based genetic diagnostic company, and trios from Radboud University Medical Center (RUMC). I identify 28 genes that were not previously robustly associated with DDs and explore how these genes differ from those that were previously known. I also develop a model-based approach to explore the likely properties of currently undiscovered genes which can inform future directions in the field. Collectively, these results reveal important insights into sources of variation in germline mutation rates as well as in mutation type. This can inform how germline mutations arise and further improve our ability to assess their contribution to rare genetic disease.