What is DeepChain™?

DeepChain™ helps biotech companies and individuals with the design of new protein molecules to attain novel activity, behavior, or purpose, and to advance basic understanding of protein function. DeepChain achieves that by making calculated variants of a known protein structure and its sequence.

What does DeepChain™ do?

DeepChain™ accelerates Protein research by offering users easy and quick access to:

  • Language Model Playground: Explore your sequences with language models trained on billions of Amino-Acids in minutes. Our transformers – state of the art machine learning models – are trained on the whole proteome. They can help uncover promising new insights from amino acid sequences.
  • AI Designer: Select a range of positions to mutate from a PDB file and a max number of mutations. Our fully autonomous, cloud-scalable AI Designer will select and deliver variants with a favourable binding energy profile in less than 24 hours.
  • MD simulations: Cross-validate your favourite variants in a few hours with full scale, GPU-powered molecular dynamics simulations.
  • Jupyter Notebooks As A Service: You can explore and analyse all your structures in hosted jupyter notebooks, with the help of any python or R library you want.
  • Patent Search module: provides information on the level of novelty of the new sequences discovered.
How does DeepChain™ work?
  1. Upload your own .PDB protein complex structure to DeepChain™ or choose one from the Protein Data Bank
  2. Evaluate possible sequence mutations using the DeepChain™ Playground – powered by Transformers trained on all known protein sequences
  3. Choose residues to mutate and let the DeepChain™ AI Designer propose mutations that minimize the binding energy of the interface – powered by InstaDeep’s Decision Making AI
  4. Cross-validate your results with DeepChain™’s Molecular Dynamics simulations and visualise them in the Protein Design Studio – powered by DGX A100 GPUs
  5. Perform further analysis on the resulting structures with Python Jupyter Notebooks. Feel free to customize your environment, this is your workplace which we will save for subsequent use as well.
Will my data be shared with any third-parties?

InstaDeep will never share or sell any of your data to any outside entity.

Why does it take 12 hours to create good designs? Is something wrong?

DeepChain™ is truly an innovative platform. To get results, it actually needs to spend time learning how to find the best sequence.

What file formats are supported?

At the moment DeepChain™ supports the PDB file format for structures.

What algorithm does DeepChain™ use to search for variants?

DeepChain™ has many different search algorithms, based on reinforcement learning, language models and genetic algorithms.

What do I do if my playground language model inference fails?

Sometimes the server might be overloaded and you might need some patience, there is a restart button.

What is AI-directed Mutagenesis?

An AI-Directed mutagenesis experiment allows one to choose a selection of mutable interface residues and let DeepChain™ propose variants that maximize the stability of the complex.

What is Site-directed Mutagenesis?

A site-directed mutagenesis experiment allows one to evaluate a given set of mutations against a variety of binding and stability related metrics.

What is the molecular dynamics simulation based on?

MD Simulations use GROMACS, an open source molecular dynamics library.

Molecular Dynamics Simulations are running too slowly, what can I do?

Sometimes MD simulations take a very long time to compute. However if you do not get results back in 24h please contact us.

I want to do something more advanced than a single interface optimization, what do I do?

For the moment DeepChain™ optimises simple protein interfaces, for positive binding. The best thing to do is to contact us.

My question is not here, what should I do?

Feel free to contact us. We will be more than happy to help!