AI and LLM-Driven Drug Discovery to Predictive Medicine of Demethoxycurcumin and Bisdemethoxycurcumin in Oral Squamous Cell Carcinoma Using Swalife Research Platform

Abstract

Oral squamous cell carcinoma (OSCC) burdens India with ~144,000 annual cases driven by tobacco/betel-induced NF-κB inflammation, PI3K-Akt survival, and MAPK invasion pathways. This 12-week computational internship at Swalife Biotech systematically addressed the research gap: no structured AI-network pharmacology workflow existed for prioritizing Curcuma longa bioactives in OSCC.
Objective: Deploy AI/LLM-driven (DeepSeek, Perplexity) 10-module pipeline to generate mechanistic hypotheses and prioritize leads from literature data.
Methods: Sequential workflow—(1) literature mining (curcuminoids/terpenoids), (2) LLM comparison, (3-4) target/lead identification, (5-10) mechanistic/clinical/PV/market/IPR/predictive modules—deployed ~120 structured prompts mapping 132+ compound-target-pathway interactions.
Results: Demethoxycurcumin (DMC; IC50 ~15μM, Akt/mTOR/PD-L1) and bisdemethoxycurcumin (BDMC; MMP2/9-JNK/p38) emerged as leads covering 82% OSCC pathways. Literature confirms apoptosis (4NQO- mice tumor reduction), stability advantages over curcumin, and mucositis adjunct potential (meta-analysis p=0.03).
Generated Hypotheses: (1) DMC/BDMC nanoformulations enhance OSCC chemoprevention (4NQO trials); (2) NF-κB/EMT synergy with cisplatin (CAL-27 spheroids); (3) Phase I window-of-opportunity safety pre-surgery.
Conclusion: Swalife's AI-pipeline demonstrates computational hypothesis generation—not experimental
validation—streamlining herbal complexity to preclinical readiness. This reproducible workflow bridges network pharmacology to predictive medicine, warranting wet-lab confirmation of prioritized DMC/BDMC as multi-target OSCC candidates.