You can find all our code on our lab GitHub page.

**PocketMiner**

PocketMiner is a neural network algorithm for predicting if/where cryptic pockets are likely to form from a single protein structure. This web-server lets you upload a PDB file and within seconds will return a new PDB file where the B-factor for each residue now gives the probability that residue participates in a cryptic pocket.

You can find all the training and validation data, as well as the model weights, here.

**FAST**

Our goal-oriented adaptive sampling algorithm.

**DiffNets**

A neural network architecture for comparing structural ensembles to understand what structural features distinguish them in a manner that accounts for overlap between the ensembles.

**Enspara**

Enspara is our open-source library for analyzing extremely large Molecular Dynamics datasets (i.e. hundreds of GB, multiple TB, etc.) Enspara contains scalable data structures and parallelized functions for construction/analysis of Markov State Models (MSMs), and an implementation of our method for assessing Correlations of All Rotameric and Dynamical States (CARDS).

**MSMBuilder**

MSMBuilder is an open source software package for automating the construction and analysis of Markov state models (MSMs). It is primarily written in the python programming language (optimized C extensions are used where appropriate to accelerate the calculations).

MSMs are a powerful means of modeling the structure and dynamics of molecular systems, like proteins. An MSM is essentially a map of the conformational space a molecule explores. Such models consist of a set of states and a matrix of transition probabilities (or, equivalently, transition rates) between each pair of states. Intuitively, the states in an MSM can be thought of as corresponding to local minima in the free energy landscape that ultimately determines a molecule’s structure and dynamics.

**And more!**

You can find our lab GitHub repository here.