Jacqueline Lam at ARDD2026: DeepDrugX for Alzheimer's Disease Drug Discovery

Jacqueline Lam at ARDD2026: DeepDrugX for Alzheimer's Disease Drug Discovery

Title history
  • Jacqueline Lam at ARDD2026: DeepDrugX for Alzheimer's Disease Drug Discovery — observed on 20261112-1930 UTC (YouTube)
🎙 Jacqueline Lam 👥 9K 📅 November 12, 2026 ⏱ 21 min 👁 3 📄 original study 🧭 2026-11-12
Available in: English (current) Français

Keywords

causal inferencedrug discoveryAlzheimer'sgraph neural networkUK Biobank

Summary

Jacqueline Lam presents DeepDrugX, a causal AI platform designed to accelerate Alzheimer’s disease drug discovery by repurposing approved drugs. The platform starts from 107 Alzheimer’s-associated GWAS genes and integrates multimodal data to construct a signed directed graph. A graph neural network learns embeddings and achieves a link prediction AUC of 0.98. The top 50 approved drugs are tested against real-world electronic health records from the UK Biobank (over 500,000 people, including 4,400 Alzheimer’s cases). After adjusting for confounders, three drugs—two blood pressure medications and a statin—are associated with lower odds of Alzheimer’s. The talk emphasizes a backward approach to identify upstream ‘bottleneck genes’ where multiple pathways converge, and uses entropy to gauge causal certainty. Independent groups have tested 11 of the top 20 candidates from an earlier version, with four entering human trials. The presentation also covers safety validations, including ranking amyloid antibodies with known safety concerns near the bottom, and results in zebrafish models. The roadmap proposes B2B partnerships and licensing for commercialisation, with human validation integrated at every stage.

171 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the information lies in its innovative combination of causal AI, graph neural networks, and real-world biobank validation to prioritize drug repurposing candidates for Alzheimer’s. The argumentation is solid: the platform uses a reverse pipeline from drugs to upstream genes, employs entropy for causal certainty, and validates outcomes in human data. The reported high AUC and independent testing by other groups strengthen the credibility of the claims. However, the presentation is promotional in tone, mentioning awards and business plans, which may dilute the purely scientific value. The argumentation is logical and consistent, but the lack of peer-reviewed publication and reliance on self-reported metrics weakens the overall argument.

Scientific Rigor, Source Quality, Title Accuracy

Scientific rigor is moderate. The talk describes a rigorous methodology with clear steps, but no detailed data or statistical analysis is shown. Sources cited include the conference website and LinkedIn, but no specific scientific publications are mentioned. The title accurately reflects the content. No comments were provided, so public reception cannot be assessed.

177 words

Title / Content Match

The title accurately reflects the content: a presentation about the DeepDrugX platform for Alzheimer's disease drug discovery.

Quality & Reliability

7/10

The talk presents a well-structured and detailed methodology, including a graph neural network with reported AUC of 0.98 and real-world validation in UK Biobank. However, results are self-reported and not yet published in a peer-reviewed venue, and the presentation includes promotional business elements.

Key Moments

Cited Sources

Concurring Sources

  • UK Biobank — Real-world data source used for validation, mentioned as a key resource

Contribution & Novelties

The main novelty is the combination of a causal AI platform with a directional graph and entropy-based certainty, allowing the identification of upstream bottleneck genes and prioritization of approved drugs for repurposing. The integration of real-world human data (UK Biobank) as a validation step before animal models is a significant contribution, addressing common pitfalls in drug discovery. The reported AUC of 0.98 and independent testing of candidates by external groups add credibility.

Pour aller plus loin :

  • Graph neural network — Core technology used for link prediction and drug ranking.
  • Causal inference — Framework for determining causal relationships from observational data, relevant to the platform’s approach.
  • Drug repurposing — Strategy to find new uses for approved drugs, central to the presentation.

121 words

Radar Profile

The profile shows strong technical depth (high scores in technical level and quality of information) with moderate overall reliability due to lack of peer-reviewed publication. The balanced scores suggest a well-presented but not yet fully verified study.

Reliability 7/10