Production of soluble, purifiable domains or multi-domain fragments of proteins is a prerequisite for structural biology and other applications. When target sequences are poorly annotated, or when there are few similar sequences available for alignments, identification of domains can be problematic. A method called expression of soluble proteins by random incremental truncation (ESPRIT) addresses this problem by high-throughput automated screening of tens of thousands of enzymatically truncated gene fragments. Rare soluble constructs are identified by experimental screening, and the boundaries revealed by DNA sequencing.