Many diseases with significant health burden and healthcare costs have been neglected in biomedical research, partly due to a lack of data. Here, we systematically harmonized >12 million primary care and hospitalisation health records from ~440,000 UK Biobank participants into 1445 collated disease terms for genetic analyses. This included ~200 diseases with >10,000 cases that are predominantly managed in primary care, like skin and non-serious infectious diseases. We quantify the heritability for these common diseases, identify novel loci with extreme effect sizes, and highlight novel roles of poorly characterised genes, e.g.PNLIPRP3as a sebaceous gland cell-specific gene for rosacea. We characterise the substantial regional pleiotropy with more than 300 independent genomic regions associated with multiple, often seemingly unrelated diseases and use these insights to generate a pan-genome disease network of shared disease loci to prioritise potential pathways contributing to multiple diseases and onset of multimorbidity. We demonstrate the value of primary care data to improve and guide genetic approaches for drug selection, repurposing, and adverse event prediction, e.g.,IGFR1as a putative multi-disease target for, among others, gout and atrial fibrillation, through network augmentation and integration of molecular quantitative trait loci. We make all results publicly available via an interactive webserver (https://www.omicscience.org/apps/phecodes). Our results provide new insights for diseases across diverse clinical specialties and provide a resource for drug discovery and mechanistic understanding.