AI for Drug Discovery #3 /DK BioAI Episode #3 ---- Live Coding: RAG for Drug Discovery

Описание: In this live coding session, I build a complete Retrieval-Augmented Generation (RAG)
system for biomedical Q&A — from raw PubMed abstracts to an AI that answers your
questions and cites the exact papers it used.

No hallucinations. Every answer is traceable to a PMID.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔬 WHAT WE BUILD (6-Step Pipeline)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
① FETCH — Pull abstracts from PubMed via Entrez API
② CHUNK — Split text with RecursiveCharacterTextSplitter
③ EMBED — Convert chunks to 384-dim vectors (all-MiniLM-L6-v2)
④ STORE — Index into ChromaDB vector database
⑤ SEARCH — Semantic similarity retrieval with QC threshold
⑥ ANSWER — Claude API generates evidence-grounded responses with PMID citations

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🧬 DEMO QUERY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
"What genes drive resistance in liver fibrosis treatment?"
→ AI returns TGF-β1, PDGFR, COL1A1 — each backed by a real PubMed paper.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📚 DK BioAI SERIES RECAP
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
EP1 — Fine-tuning PubMedBERT for biomedical relation extraction + RAG integration
EP2 — Graph Neural Network (GNN) for drug-target interaction prediction (AUC 0.87)
EP3 — Live coding: full RAG Q&A pipeline from scratch ← YOU ARE HERE

Three AI technologies. One integrated drug discovery pipeline.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
💻 CODE & RESOURCES
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GitHub: github.com/kdh4win4
ORCID: 0000-0002-5794-6222

Tools used: Python · BioPython (Entrez) · LangChain · ChromaDB ·
SentenceTransformers · Claude API (Anthropic)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔔 If you're in biomedical AI, drug discovery, or computational biology —
subscribe for more hands-on builds.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Dohoon Kim, M.S. | Computational Biologist | AI for Drug Discovery



#RAG #RetrievalAugmentedGeneration #DrugDiscovery #BiomedicalAI #LLM
#PubMed #ChromaDB #SentenceTransformers #LangChain #ClaudeAPI
#ComputationalBiology #BioinformaticsAI #PubMedBERT #VectorDatabase
#AIforScience #LiveCoding #PythonAI #DKBioAI #MachineLearning
#NaturalLanguageProcessing

#人工知能創薬 #バイオインフォマティクス #医療AI
#인공지능신약개발 #바이오인포마틱스 #의료AI #생물정보학
#人工智能药物发现 #生物信息学 #医疗AI
#KünstlicheIntelligenz #Bioinformatik #KIMedizin
#InteligenciaArtificial #Bioinformática #IAMedica
#IAMédicale #Bioinformatique #DécouverteDeMédicaments
#الذكاءالاصطناعي #بيوانفورماتيكس
#AIdellaSalute #Bioinformatica
#yapayzekaileIlaçKeşfi #biyoenformatik
#AInaOtkrytieLekars #биоинформатика