Gene Therapy AAV Clinical Trial Pre-screening: Citation and

Data Characteristics

Gene therapy AAV (adeno-associated virus) clinical trial pre-screening data primarily originates from global clinical trial registries (e.g., ClinicalTrials.gov, EU Clinical Trials Register), academic journals, conference papers, and patent databases. Update frequencies vary; clinical trial registration information typically updates upon trial initiation, modification, or results publication, ranging from weeks to months. Document structures are mostly structured or semi-structured. They include trial protocol descriptions, subject inclusion/exclusion criteria, drug information, research institutions, investigators, efficacy endpoints, and adverse event reports. Fields and units are highly specialized. Examples include AAV serotype (e.g., AAV2, AAVrh.10), viral vector dose (e.g., vg/kg), gene expression vector name (e.g., CMV-GFP), disease indication (e.g., SMA, DMD), and route of administration (e.g., intravenous injection, intrathecal injection). Some data exists as detailed research reports or supplementary materials in PDF format.

Constraints on Citation and Traceability

The specialized and diverse nature of AAV gene therapy clinical trial data imposes specific requirements on citation and traceability. First, the broad range of data sources—registries, academic papers, and patents—requires the system to integrate multi-source heterogeneous information and accurately identify the original source of each information fragment. Second, inconsistent data updates, especially dynamic changes in clinical trial status, mean traceability must consider information timeliness to avoid citing outdated or withdrawn data. The coexistence of structured and semi-structured data necessitates effective parsing for different document formats to accurately identify core fields like AAV serotype and viral vector dose. Furthermore, specialized medical terminology and biological units, such as vg/kg, require the system to correctly interpret their meaning during citation and prevent misunderstandings due to unit confusion. The ability to parse detailed PDF reports determines whether critical traceability information, such as adverse event rates, can be accurately extracted from unstructured text.

Configuration Settings

Configuration Item	Suggested Value	Rationale
`Chunk Length`	`500–800 characters`	Balances the completeness of AAV gene therapy clinical trial descriptions with model processing window limitations.
`Recall Count`	`10–15 entries`	Covers multiple dimensions of information, including clinical trial protocols, indications, and adverse events, ensuring comprehensiveness.
`Similarity Threshold`	`0.75–0.85`	Filters out irrelevant trial information while retaining potentially relevant AAV types or disease matches.
`Rerank Return Count`	`5 entries`	Prioritizes displaying information most relevant to the query, such as AAV vectors, target genes, or disease indications.
`Max Reference Token Count`	`3000 tokens`	Ensures sufficient capacity for key paragraphs from multiple clinical trials, facilitating model context understanding.
`Max PDF Pages to Parse`	`50 pages`	Accommodates detailed clinical research reports or trial protocols, preventing information loss due to page limits.

Common Pitfalls

Cited clinical trial information in the output does not match the original document. This happens when units for critical fields like viral vector dose are not correctly identified during document preprocessing, leading to numerical parsing errors.
Model responses cite AAV clinical trial data that has been terminated or withdrawn. This occurs when the knowledge base update mechanism fails to timely synchronize the latest status from registries like ClinicalTrials.gov.
Under complex queries, such as those involving comparisons of multiple AAV serotypes or different routes of administration, the number of recalled reference entries is too low. This is because the Recall Count is configured too low, failing to cover sufficient diverse information.

Verification Steps

Select multiple representative AAV gene therapy clinical trial queries. Check if the Trial ID cited in the output exactly matches the NCT number or EudraCT number in the original knowledge base source.
For knowledge base entries containing PDF reports, verify specific data cited in the model's response, such as adverse event rates, by comparing them with corresponding values in the original PDF file to confirm accurate parsing.
Design complex queries involving different AAV serotypes, disease indications, and routes of administration. Check if the recalled reference entries cover all relevant dimensions and evaluate the recall precision corresponding to the Similarity Threshold.

The values provided are common starting points. Measure them against your own samples.

Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-21.