Graduate student
Neutron reflectometry (NR) is one of the techniques of choice to fill the data gap for disease-relevant proteins or peptides at cell membranes under pharmaceutically relevant conditions. A NR-based innovative measurement approach, called the Reflectometry-driven Optimization And Discovery of Membrane Active Peptides (ROADMAP), is currently being developed at NIST to create an autonomous biomolecule design infrastructure. Antimicrobial peptides (AMP), which can efficiently disrupt bacteria membrane upon interaction with the lipid bilayer, are the focus of ROADMAP.
The successful candidate will design and implement the machine learning (ML) and bioinformatics component of ROADMAP, which will determine the sequence of NR measurements leading to a comprehensive dataset within the experimental time and resource constraints. The main tasks of the recipient will be: to develop data structures that capture AMP properties, multimodal NR observables and experimental conditions; to select appropriate ML architectures; to develop molecular modeling and simulation tools for the structural and dynamic characterization of AMPs; to interface the molecular modeling and simulation tools with the ML ones; to validate the resulting framework on AMPs of interest in the context of ROADMAP.
Machine Learning and Molecular Modeling for Neutron Reflectometry
- A Ph.D. in computer science, computational biology, statistics, mathematics, or a related field.
- 3+ years of experience in machine learning, preferably but not necessarily with application to biology or chemistry.
- Familiarity with Pytorch and Tensorflow platforms for ML modeling.
- Familiarity with Pytorch Geometric platform for building and evaluating Graph Neural Networks.
- Familiarity with RDkit and pysmiles platforms for processing and analyzing molecular sequences.
- Familiarity with molecular modeling and simulation tools such as NAMD, Gromacs, VMD, and CHARMM.
- Ability to develop prototypes of tools needed to analyze data.
- Strong oral and written communication skills.
Key responsibilities will include but are not limited to:
- Assess AI models for computer-aided drug design.
- Define data structures for the representation of biomolecular entities of interest.
- Design and implement ML models for the prediction of peptide antimicrobial properties and assess their prediction accuracy. The ML models should be generative (i.e., generate novel peptides with optimal properties).
- Study the peptide representation associated to the implemented ML models, then define ML-specific metrics to characterize the prediction performance. Based on the collected evidence, provide guidance for the next batch of NR measurements, which will yield additional training data in an iterative fashion.
- Assess currently available molecular modeling and simulation tools for peptide structural and dynamic characterization. Focus specifically on ML tools enabling the representation of the peptide conformational families.
- Based on the above findings, design ML tools for the efficient sampling of peptide conformational ensemble both in solution and in lipid membranes.
- Integrate the structural and dynamic insight from ML-based sampling with the above-mentioned generative ML models.
- Write manuscripts describing the above work.