PDAOAI
AI-augmented literature intelligence for precision oncology research
Overview
PDAOAI leverages large-scale natural-language processing to accelerate hypothesis generation and biomarker discovery. Built on transformers, vector databases, and clustering algorithms, the platform converts Pubmed Research Papers into an interactive knowledge graph that answers domain-specific questions in real time.
In the question-answering session, the user query was transformed into an embedding vector, and then an appropriate similarity measure (e.g., cosine distance) was used to identify the embedded abstract vectors closest to the embedded query vector: the abstracts corresponding to these matching vectors were then fed to the question-answering model (in the form of context), along with the query, to arrive at an answer to the query.
— From the paper
End-to-End Workflow
1. Ingestion
- Keyword-driven PubMed retrieval
- Automatic PDF to text conversion
- Metadata enrichment (DOI, MeSH)
2. Semantic Indexing
- Snippet embeddings via OpenAI text embedding 3 large
- Vector storage in Qdrant
- Noise reduction with HDBSCAN clustering
3. Retrieval & QA
- Cosine-similar nearest-neighbor search
- Context assembly & prompt templating
- LLM reasoning & answer synthesis
4. Front-End Delivery
- React + Material-UI dashboard
- Rich tables, charts & export tools
- Node.js REST API layer
Peer-Reviewed Publications
1. Positive Prognostic Overall Survival Impacts of Methylated TGFB2 and MGMT in Adult Glioblastoma Patients
Cancers 2025, 17, 1122. DOI: 10.3390/cancers17071122
This study employed the Chatbot pipeline to screen 3,000+ GBM-related abstracts, guiding the multivariate survival analyses that uncovered TGFB2 methylation as a superior prognostic marker. Read Paper
2. TGFB2 Gene Methylation Drives Positive Prognostic Responses in Pancreatic Ductal Adenocarcinoma
International Journal of Molecular Sciences 2025, 26, 5567. DOI: 10.3390/ijms26125567
The Chatbot’s AI-augmented literature triage accelerated biomarker discovery in PDAC, enabling integrative methylation and immune-infiltration analyses across 1,195 abstracts. Read Paper
Get in Touch
Interested in deploying PDAOAI or collaborating on oncology data science? Reach out at ir@oncotelic.com.