Computational Biology & AI

AI-Driven Protein Engineering and Molecular Simulations

X-Tour is interested in machine learning models and molecular simulation pipelines for protein engineering and computational biology research.

1
Curate
2
Train
3
Simulate
4
Predict
Scroll to explore

Sequence & Genomics

Machine learning that looks for patterns in biological sequences, both protein and genomic

Molecular Dynamics

Simulations that show how molecules move and change over time

Structure Analysis

Tools that predict the structure of a protein, and what that structure tells us

Our Areas of Interest

Prediction Meets Simulation

Two fields we are interested in: machine learning for protein properties, and physics-based molecular simulation.

Predictive Modeling

ML for Protein Properties

Machine learning applied to protein properties: working out how a protein will behave from its sequence and structure alone. What interests us is how much a model can actually learn from well-prepared data, how published methods hold up on data they have not seen, and how to compare a prediction fairly against a result from the lab.

  • Predicting how a protein behaves from its sequence and structure
  • Learning from large collections of biological sequences
  • Well-prepared data and fair comparisons
  • Knowing when a prediction should not be trusted
Molecular Simulations

Simulating Molecular Behaviour

Molecular simulation shows how molecules move and interact with each other, one atom at a time. What interests us is how much a simulation can explain about what a statistical model merely predicts, and how much computing that costs.

  • How molecules move and change their structure
  • Reaching events that happen only rarely
  • The physics a simulation rests on
  • How strongly molecules bind to each other
Technology

The Pipeline We're Interested In

An end-to-end workflow, from raw biological data to validated predictions, combining ML and physics-based simulation.

1

Data Curation

Sequences, structures, and measurements gathered from public databases and from partners, then checked carefully for quality.

2

Model Development

Models trained on biological data from one specific field, then tested on data they have never seen.

3

Simulation & Validation

Simulation gives an independent check on what a model predicts, from the physics side. Predictions compared against results from the lab.

4

Deployment

Finished tools and simulations that a research team can use through an API, or fit into the way it already works.

Our Approach

Where ML Meets Biophysics

We start from the biology, not the algorithm. Whether the question is a predictive model or a microsecond-scale simulation, we want it grounded in the underlying biophysics.

01

Research-Driven

Every model that interests us starts with the biophysics. We study the underlying physics and the molecular mechanisms before considering any computational approach.

02

Physics-Informed

The pattern recognition power of machine learning combined with the physical principles that govern how molecules actually behave. What interests us is whether that combination gives models which generalize beyond the training data.

03

Aimed at Real Use

We are interested in tools that would be practical to use: reproducible pipelines, well-documented APIs, and systematic benchmarking against experimental results.

Latest Insights

Research & Perspectives

Exploring the intersection of AI, molecular simulation, and protein engineering.

Articles coming soon.