ROBUST

Molecular dynamics simulations are well suited for studying molecular recognition. A key challenge is the derivation of meaningful descriptors from the raw MD coordinates.

ROBUST provides a set of tools to calculate physics based descriptors from molecular dynamics simulations.

Along with the transformers we provide a number of examples cases.

Requirements

• python ≥ 3.6
• numpy
• pandas
• scipy
• scikit-learn
• voronota
• Schrodinger ≥ v.18.1

The Schrodinger Software Suite is required to process molecular dynamic simulations.

For examples, the descriptors have been precomputed thus the examples do not require Schrodinger.

Transformers

A series of transformers to calculate and parse descriptors from molecular dynamics simulations. Transformers were developed for use with Schrodingers Protein-Ligand Database however, they can also be called from the command line or imported in python.

Usage

From command line

All transformers can be run from the commandline; For a list of arguments simply run:

$SCHRODINGER/run python transformer.py --help

As a python module

All transformers can be imported into a python script or jupyter notebook. An example is shown in the examples/DHFR/calculate_descriptors.ipynb .

From the PLDB

All transformers can be incorporated into a PLDB pipeline. For details on our in-house analysis pipeline contact Judy Huang.

Examples

HIV

Development of a predictive model for inhibitor affinity of drug resistant HIV-1 protease variants using pMD and ROBUST

Referece paper: Deciphering Complex Mechanisms of Resistance and Loss of Potency through Coupled Molecular Dynamics and Machine Learning

DHFR

Application of pMD and ROBUST to identify Trimethoprim resistant dihydrofolate reductase variants

Reference paper:Deciphering Antifungal Drug Resistance in Pneumocystis jirovecii DHFR with Molecular Dynamics and Machine Learning

Contact

Florian Leidner:

florian.leidner@umassmed.edu

Judy Huang

qiuyu.huang@umassmed.edu

Celia Schiffer:

celia.schiffer@umassmed.edu