Open-source tools for early drug discovery, chemistry, and the life sciences.
Zero-configuration Named Entity Recognition for drug discovery, chemistry, and biological sciences.
Extracts compounds, targets, genes, diseases, bioactivities, and assays from scientific text using two complementary engines:
- NERExtractor — LLM-powered extraction via Gemini or local Ollama models, with contextual understanding for novel entities
- FastNERExtractor — Dictionary-based extraction using YAML gazetteers with exact, regex, and fuzzy matching (~393ms per abstract)
fromstructflo.nerimportNERExtractorextractor=NERExtractor()
results=extractor.extract(abstract_text, profile="CHEMISTRY")Chemical Structure Entity Recognition A Deep Learning detector for identifying compound panels and their matching labels/IDs in scientific documents.
Detects bounding boxes containing rendered chemical structures and their associated label IDs (e.g. CHEMBL12345) across PDFs and images and full-page documents.
fromstructflo.cser.pipelineimportChemPipelinepipeline=ChemPipeline()
results=pipeline.run("paper.pdf")Most structflo libraries are released under the Apache License 2.0. Please review individual repos for any exceptions.