SOW The MD Simulation Platform.docx

Statement of Work (SOW)/Performance Work Statement (PWS): Molecular Modeling Simulation Platform or comparable

The Office of Product Quality and Research (OPQR) conducts research to advance the adoption of emerging technologies for pharmaceutical manufacturing and support the regulatory review, inspection of FDA regulated products. OPQR uses artificial intelligence (AI) and machine learning (ML) based advanced molecular dynamic simulation tools for assessing the impact of manufacturing process parameters, material attributes and end user handling on the quality and performance of complex drug products and biologics. These studies include but are not limited to atomic and molecular level modeling and simulation of protein-excipient interaction in liquid, frozen and dried products, drug-interface interaction, liposome and nanomaterials structural dynamics, drug-excipient interactions, etc. To this end OPQR is seeking to extend the license for the current molecular simulation platform or comparable platform and associated technical support to support ongoing research.

The MD simulation platform shall provide a comprehensive workflow, data management, and storage management, computing resource management capability. The platform needs to enable parallel simulation of at least two MD simulation projects. The platform and technical service providers shall have a proven track record on solving pharmaceutical and regulatory science challenges involving molecular dynamics and simulation. This platform will be deployed on FDA High-performance workstation (Precision 7920 Tower) with a Linux based system that meets the required specifications.

As the system will be installed in FDA DPQR proprietary High-performance workstation (Precision 7920 Tower), there is no security risk. Technical risks on knowledge transfer are mitigated via training and support packages included. Administrative and financial risks are mitigated via annual renewal of the subscription contract.

Item 1 : Molecular Dynamic Modeling and Simulation Platform

The required technical features of molecular dynamic simulation software package are:

· Shall allow seamless transfer of current workflow and projects without the need to rebuild and validate current models

· Shall be a high-performance molecular dynamics simulations platform capable of designing, modeling and simulating biological interactions, antisense oligonucleotides, protein-excipient and protein-ice interactions, protein-ligand complexes, viral capsid and macromolecular complexes with high throughput and scientific accuracy.

· Shall include an all-atom ad coarse grain model

· Shall include a machine learning model for predicting the interactions between formulation components and biologics or small molecule active pharmaceutical ingredients

· Shall employ an accurate and scalable machine learning force fields with broad molecular coverage to describe the interactions between atoms and predict relative energies

· Shall have a validated, proven, stable and robust framework for the calculation of energies and forces for atomistic force field models.

· Shall describe interactions between atoms in system using modern force fields with comprehensive coverage of chemical space for small molecules, biologics and materials science applications.

· Shall be a professionally managed software that is broadly available and used by the pharmaceutical industry to facilitate knowledge transfer from and to the FDA.

· Shall leverage machine learning to accurately and rapidly predict the aqueous phase pKa values and protonation state distributions of monoclonal antibodies and globular proteins.

· Shall employ supervised machine learning (ML) for predicting molecular properties based on chemical structure.

· The supervised learning platform shall be easy to implement by novice and experienced users.

· Shall be a user-friendly machine learning tool for relating properties derived from molecular simulations to key experimental properties.

· Shall predict quantitative structure activity/property relationship models.

· Shall offer a robust workflow for data management, storage, and computing resource management to streamline simulation projects.

· Shall manage and stores large volumes of image and quantitative data efficiently.

· Shall perform stability and reactivity analysis and facilitates the study of biologics and small molecule drug stability and reactivity under various conditions.

· Shall be a fast, general-purpose computing on graphics processing units (GPGPU) enabled molecular dynamics simulation program.

· Shall simulate systematic temperature variation and the progressive removal of water from a complex model system.

· Shall include well validated workflows for protein structure preparation including homology modeling particularly for antibodies, protein loop sampling, structural relaxation and protonation state assignment.

· Shall simulate and analyzes protein-excipient combination and identify excipient hotspots near protein surfaces in solutions and freeze-dried products.

· Shall simulate protein excipient interactions of not less than 70000 atoms over a simulation duration of not less than 270 ns

· Shall support three-dimensional visualization of reactions and analysis.

· Shall include analysis tools for protein structures sampled during simulations of protein containing systems such as secondary structure analysis, protein patch analysis.

· Shall be capable of constructing systems containing a mixture of proteins with complex composition of non-biological components.

· Shall characterize complex mixtures of components including radial distribution functions, spatial density analysis, glass transition temperatures and diffusion rates.

· Shall conduct simulations in NPT or NP𝝻T ensembles (constant temperature and pressure), with typical temperatures of 233.15 K and 300 K, or other mutually agreed-upon values.

· Shall simulate interactions between formulations and lipid-based systems, liposomes, or other nanomaterials.

· Shall analyzes the 3D distribution of components within lyophilized formulations, including small and complex molecules.

· Shall include parameter analysis including but not limited to trajectories, root mean square deviation (RMSD), energy profiles, and other critical data from simulation systems.

· The MD simulation platform shall have a user accessible script-based automation of complex modeling workflows.

· The above features shall be available in a single, integrated suite of software with a well-developed graphical user interface for visualizing molecular systems, enable interpretation of results and facile interoperability of the protein-focused and materials science-focused capabilities for molecular dynamics and machine learning.

· Shall be deployable on DPQR network. and compatible with hardware in OPQR Precision 7920 Tower:

· Memory: 128GB

· GPU: Dual Nvidia RTX A5000 (8192 CUDA cores each), 24GB

· Operation System: Ubuntu

· CPU: Intel Xeon Gold 5220R 2.2GHz (24 cores)

· Storage: 512GB

· Shall accelerate performance of above Graphics processing unit (GPU) or comparable by at least 60%

· Alternatively, the vendor may supply software with compatible government approved hardware, GPUs or high-performance computing platform.

· The platform and the vendor’s scientists shall be well published and have a proven track record on solving pharmaceutical and regulatory science challenges involving MD simulation.

· Shall provide technical support and perform simulation parallel to FDA’s projects at client requests.

· Shall including packages, license, and libraries for:

· Machine learning-based predictive modeling.

· Molecular visualization and chemical informatics.

· Conformational sampling.

· Scaffold replacement for novel molecule design.

· pKa and protonation state prediction tool.

· 3D-QSAR modeling.

· Ligand-receptor docking.

· Ligand preparation and stereochemical variation.

· Molecular mechanics simulations.

· Predicts drug permeability across membranes and interfaces.

· Protein structure prediction and refinement.

· Quantum mechanics and Molecular mechanics simulations for reactions of proteins and ligands.

· Binding site identification for proteins.

· Design, characterization, and optimization of drug formulations and delivery systems at the molecular level.

· Biologics modeling (e.g., antibodies).

· KNIME, Maestro, and OPLtS Access Licenses: Data workflows, GUI, and force field support

Item 2: Delivery, Installation, On-site training and Technical Support The vendor will cover all costs associated with the following:

· System software, shipping, and delivery.

· Full system installation in person or screen sharing support on FDA OPQR in house high-performance workstation with the help of client’s IT professional.

· Provide all necessary training to run the projects on simulating, analyzing and visualizing the formulations, molecular interactions, and manufacturing of complex drugs.

· All necessary training and technical support to execute associated projects on simulating, analyzing and visualizing of formulations, interfacial and molecular interactions, and manufacturing of complex drugs.

· Shall offer a minimum of 3 days of full time employee hours for comprehensive support, services for constructing protein formulation structures and simulation with their resources parallel to FDA DPQR Precision 7920 Tower at FDA OPQR requests.

· Comprehensive modeling support including documentation, rapid response to queries and knowledge transfer to the FDA, The molecular modeling and simulation platform will be delivered to the following location:

ATTN: Maxwell Korang-Yeboah Food and Drug Administration 10903 New Hampshire Ave Building 64, Room 1016 Silver Spring, MD 20993