Session 07 · No-Code & Agentic AI for Life Sciences

Deep Learning
in Drug Discovery

Target ID · Compound Screening · AlphaFold · ADMET

Md. Jubayer Hossain · Founder & CEO
DeepBio Limited · DeepBio Academy

No Code & Agentic AI for Life Sciences · Session 07 · June 2026

Pipeline Overview

ML in Drug Discovery Pipeline

🎯
Target ID
Disease-linked gene/protein.
💊
Hit Gen
Finding initial molecules.
⚖️
Optimization
Potency & Selectivity.
🧪
ADMET
Safety & Pharmacology.

🚀 The AI Advantage

Reduces time from 10 years to < 5 years. Shifts discovery from "Trial and Error" to "Predictive Design."

📦 No-Code Stack

Open Targets, ChEMBL, SwissADME, AlphaFold, and Agentic Workflows.

Target Discovery

Target Identification: Open Targets

Open Targets Platform integrates genetics, omics, and literature data to prioritize targets.

  • Association Score: Strength of evidence linking gene to disease.
  • Tractability: Can this target be drugged with small molecules or antibodies?
  • Safety: Known target-related toxicities.

💡 Use Case

"Find all druggable kinases associated with Rheumatoid Arthritis with a genetics score > 0.5."

Chemical Intelligence

ChEMBL + Claude for Queries

ChEMBL is the world's largest open bioactivity database. Agents act as the interface.

# Agentic ChEMBL Query Prompt: "Search ChEMBL for molecules inhibiting EGFR. 1. Filter for IC50 < 10nM. 2. Return a table with SMILES, Molecule Name, and logP. 3. Highlight those that pass Lipinski's Rule of 5."

🧪 SMILES Awareness

Claude understands SMILES notation (e.g., CN1C=NC2=C1C(=O)N(C(=O)N2C)C) and can reason about functional groups.

🔍 Rapid Screening

Instead of manual SQL, use natural language to find "leads" in seconds.

Predictive Chem

Molecular Property Prediction

🏗️ Quantitative Structure-Activity (QSAR)

Predicting biological activity based on chemical structure features (fingerprints).

🛠 No-Code Tool: DataWarrior

Desktop app for visualizing chemical space and predicting druglikeness, mutagenicity, and toxicity.

"Machine learning allows us to screen billions of virtual molecules without synthesizing a single one."
Optimization

Drug Repurposing

AI finds "new tricks for old drugs." Faster route to clinical use since safety data exists.

  • Signature Matching: Finding drugs that reverse a disease's gene expression profile.
  • Network Analysis: Identifying secondary targets for approved drugs.
  • Side-Effect Analysis: Turning "side effects" into primary therapies.

🌟 Success Story

Sildenafil (Viagra): Originally for hypertension, repurposed for erectile dysfunction based on clinical observations.

Structural Biology

AlphaFold for Target Structure

DeepMind's AlphaFold has solved the 50-year-old "protein folding problem."

  • Atomic Resolution: Predicts 3D structure from amino acid sequence.
  • Binding Site ID: Identify where a molecule can actually attach.
  • Tractability: See if the protein has a "pocket" suitable for a drug.

Predicted structure ➜ Virtual Docking

Pharmacology

ADMET Property Analysis

💧 Absorption

Solubility, Permeability (Caco-2), P-gp substrate.

📦 Distribution

Plasma Protein Binding, BBB permeability.

🔄 Metabolism

CYP450 inhibition/substrate (half-life).

🚽 Excretion

Renal/Hepatic clearance paths.

☠️ Toxicity

hERG inhibition, Ames mutagenicity, Hepatotoxicity.

Agentic Orchestration

The Discovery Loop

Connecting the dots from gene to literature with a single agent directive.

  • Identify: Gene associated with Neurodegeneration (Open Targets).
  • Structure: Predict protein structure (AlphaFold Database).
  • Screen: Query ChEMBL for similar leads (Claude Agent).
  • Refine: Predict ADMET profile (SwissADME / DataWarrior).
  • Ground: Find latest literature validating the target-drug pair (PubMed).
    • Summary

    Session 07 Key Takeaways

    • AI transforms the "Hunt for the Needle" into "Needle Engineering."
    • Open Targets priorities targets using evidence scores.
    • Claude + ChEMBL provides a natural language drug search.
    • AlphaFold provides the 3D map for virtual docking.
    • ADMET filters identify early if a drug will fail in clinical.
    • Agentic workflows link data across the entire discovery arc.

    Next Session: Biomedical Image Analysis — AI vs. Radiologist accuracy and custom vision models.

    No Code & Agentic AI for Life Sciences

    Thank You

    Get in Touch

    Md. Jubayer Hossain

    bio.link/hossainlab

    © 2026 Md. Jubayer Hossain · No Code & Agentic AI for Life Sciences — Session 07