Most genetic studies of neuropsychiatric disorders focus solely on disease risk at the gene level. However, nearly all genes expressed in the brain routinely produce multiple transcript isoforms. Recent large-scale analyses have shown that examining the effect of disease variants on isoforms doubled the detection of schizophrenia risk signals, suggesting that isoforms are often the targets of disease variants. Despite this progress, we have a limited understanding of which isoforms cause disease and how they do so, which hampers efforts to uncover disease mechanisms and develop therapies. We aimed to address this gap by identifying and characterising disease-associated isoforms in schizophrenia. We systematically surveyed PubMed/Scopus to identify two isoform-resolved, whole-transcriptome resources from the developing human brain. We assembled a prenatal, trimester-aware database of isoform–trait associations (effect sizes, method, replication). Colocalization of isoform QTL signals with GWAS using TWAS, COLOC, and eCAVIAR prioritised high-confidence risk isoforms (n=2373). We identified 57 genes comprising 73 high-confidence risk isoforms that were corroborated across different methods and/or studies. Fine-mapping further refined the signal for 10 of these isoforms to a single, high-confidence variant that appears to influence both their expression and schizophrenia risk—for example, STAB1-205 (rs7612511), ABCB9-215 (rs1716183), and YWHAE-206 (rs9905529). These implicated isoforms are predicted to feature distinct proteoforms for STAB1, YWHAE, and an alternative 5′ UTR in ABCB9. Notably, AlphaFold modelling of the YWHAE risk isoform predicts the loss of 6 of 9 α-helices, disrupting the canonical 14-3-3 dimer interface—consistent with a loss of function. These findings provide precise, testable targets for experimental follow-up and demonstrate that isoform-resolved analyses reveal actionable biology often missed by gene-level approaches.