Polly Xtract

Polly Xtract is an advanced, proprietary AI-driven data extraction and curation engine developed by Elucidata, designed to intelligently extract and structure complex biomedical data from a wide array of unstructured and semi-structured sources. Built for biopharma, clinical research, and research informatics teams, it transforms dense documents — including clinical trial protocols, publications, EHR exports, and regulatory filings — into clean, structured schemas with 98% accuracy and up to 500x the speed of manual curation.

Polly Xtract serves as a core technological engine powering Elucidata's expert curation services and is available for early-access partnerships and enterprise deployments. It is purpose-built for biomedical metadata curation, drawing on six years of curatorial decision-making and quality assurance, and is trusted by leading biopharma organisations worldwide.

Key Performance Capabilities

• 98% accuracy for fully automated, high-accuracy data extraction across complex fields from any publication or document.
• 500x faster than human counterparts, capable of extracting highly accurate data from a publication in as little as one second.
• Highly scalable — supports launching thousands of parallel extraction jobs simultaneously, built for enterprise-scale document extraction and metadata enrichment.
• Achieves ≥90% accuracy for simple (binary) and complex (textual) fields, ~87% accuracy for moderate (numeric) fields, and F1 scores above 85% across most field types.
• 100% consistency across binary fields over multiple runs and 100% field-level coverage with no missing predictions.
• 90% groundedness, with extracted values traceable directly to source documents.
• Delivers a 4× increase in throughput, matching the monthly output of a three-person expert curation team.

Core Features

• AI-Generated Metadata Schema: Leverages an LLM-powered system to auto-generate the optimal metadata schema directly from document content, with no restrictive templates required.
• Bring Your Own Schema: Allows users to define exactly which fields matter by creating or importing their own ontologies, schemas, or data models for full flexibility and control.
• Any Document, Any Format: Extracts structured data from PDFs, scanned images, Word documents, spreadsheets, and audio recordings — source- and modality-agnostic.
• Transparent AI Reasoning: For complex or inferred data fields, Polly Xtract provides human-readable explanations showing how answers were derived, with structured reasoning logs and field-level evidence.

Multi-Agent System Architecture

• Document Parsing Agents: Ingest structured and unstructured sources including GEO, publications, and supplementary materials.
• Extraction Agents: Operate on text, tables, and figures to extract field-specific values.
• Ontology Mapping Agents: Normalise outputs using controlled vocabularies such as MeSH, Cell Ontology, and Disease Ontology.
• Reasoning/Validation Agents: Handle conflicts, perform plausibility checks, and reconcile outputs across sources.
• QC/Review Loop: Human-in-the-loop review for flagged fields, ensuring the highest data quality standards.
• Schema-Enforced Output: Structured data is written to Polly Atlas and exposed to downstream tools such as Polly KG.

Supported Data Types and Sources

• Textual: GEO pages, full-text publications (PMC, publisher PDFs), supplementary files (PDF, Excel, Word), trial protocols and case report forms, EHR/EMR exports, and regulatory documents.
• Tabular/Embedded: HTML tables (GEO, PMC), image-based or LaTeX tables in PDFs, and CSV/TSV files in supplementary sections.
• Unstructured/Free-text: Clinical narratives, figure captions, and journal discussions.
• Multimodal: Cross-referencing across GEO, supplements, and external links; context-sensitive extraction from multiple documents per study.

Extractable Metadata Fields

• Supports extraction across 23+ metadata fields including disease, cell type, sample type, tissue, perturbation or compound, platform/technology, organism, donor ID, and study accession.
• Handles both raw entity extraction and ontology-based normalisation (e.g., mapping "AML" to DOID:9119).
• Schema-flexible, supporting both pre-defined and user-defined metadata fieldsets.

Core Use Cases

• Omics metadata harmonisation: Extract and standardise sample-level metadata across GEO, ArrayExpress, and internal datasets.
• Clinical trial parsing: Auto-extract schema elements from protocols including arms, endpoints, and eligibility criteria.
• EHR/EMR extraction: Structure unstructured patient data for real-world evidence and cohort analytics.
• Toxicology digitisation: Transform legacy reports into structured datasets for analysis or regulatory submission.
• Scientific literature curation: Extract study design, compound information, and results from publications and supplements.
• Assay result digitisation: Normalise outputs from vendor PDFs or spreadsheets into LIMS-compatible formats.

Who Should Use Polly Xtract

• Curation teams handling omics or clinical datasets at scale.
• Informatics and R&D teams building knowledge graphs, data lakes, or FAIR repositories.
• Data scientists preparing training datasets for AI and machine learning models.
• Clinical operations or biomarker groups working with protocols and lab data.
• Pharma and diagnostics teams receiving unstructured data from partners.

Workflow: From Upload to Outcome

1. Upload any trial document, publication, or unstructured source file.
2. Auto-extract schema elements in seconds using the AI-powered multi-agent pipeline.
3. Review, refine, and export structured data with ease.

Polly Xtract is deployed on dedicated AWS compute resources within a separate VPC, with AES-256 encryption at rest and in transit, TLS/SSL, and AWS IAM for fine-grained permission management. Users retain complete ownership and control over their data, including export, editing, retention, and storage. The platform integrates with downstream Elucidata tools including Polly Atlas and Polly KG, and is available for early-access partnerships and enterprise deployments for teams managing high-volume biomedical data extraction.