← All work
Case StudyTier AResearchSep 2024

M.Sc. Thesis — Drug–Target Binding Affinity Prediction

Bioinformatics research combining deep feature extraction, similarity descriptors, and gradient boosting for drug–target binding affinity prediction.

AIDataResearch
Evidence
~10% avg improvement reported in thesis

Context

Sep 2024 · M.Sc. research · completed · Maturity: Research.

Problem

Improve predictive performance on established drug–target affinity benchmarks.

Users

Academic research context at Sharif University of Technology.

My Role

M.Sc. thesis research and implementation.

Product Decisions & Responsibilities

  • Framed evaluation around benchmark datasets and reproducible comparisons.

Technical Approach

  • Worked with Davis/KIBA data, deep feature extraction, similarity descriptors, CatBoost/gradient boosting, and benchmark evaluation.
PythonMachine learningCatBoostDeep featuresDavis/KIBA

Constraints

  • Performance varies by model, dataset, and metric.
  • Research results are not clinical claims.

Product Flow

  1. Prepare benchmark data
  2. Extract features/descriptors
  3. Train models
  4. Evaluate CI/MSE
  5. Compare baselines

Outcome

  • Approximately 10% average improvement as reported in the thesis.

What I Learned

  • Research metrics are only meaningful with the benchmark and evaluation context attached.

Evidence Boundary

What this does — and does not — prove

Use “approximately 10% average improvement as reported in the thesis”; no universal or clinical claim.

Next Iteration

Extend the product only after the next decision can be tied to stronger user, operational, or business evidence rather than adding features for their own sake.