The objective of this Statement of Need (SON) was to develop advanced computational methods and validated models capable of predicting toxicological and physical properties related to novel energetic materials (EM) including binders, plasticizers, oxidizers, and fuels. Proposals that offered a systems approach to investigate compounds and synergistic effects of their formulations were preferred. Work proposed against this SON should have creatively leveraged EM datasets, public datasets, and insights from past property prediction work to create new representations and models for properties of interest to SERDP. Models meeting the requirements of this solicitation should have been capable of predicting [one or more of] the following toxicology parameters on energetic material datasets:
Mutagenicity
Acute and chronic toxicity using representative exposure routes
Oral, dermal, and inhalation
Aquatic, terrestrial and avian
Irritation/sensitization
Developmental/reproductive toxicity
Carcinogenicity
Additionally, models were needed for prediction of physical properties relevant for assessing environmental fate and transport, including the following parameters:
Aqueous solubility
pH sensitivity
Octanol-water partition coefficients
Vapor pressure/Henry’s Law Constant
Biodegradability in soil or fresh water
Photolytic stability
Models with clearly defined domains of applicability and that were interpretable were preferred. Proposers could have also included experimental validation of models with energetic materials or structural analogs, and the creation or curation of experimental datasets for use with their proposed method. Proposals should have discussed validation metrics and methods for splitting of datasets that would be used for model development. It was also essential that new modeling techniques were user-friendly to the risk assessment community. Ideally, models should have been available for use by energetics researchers for incorporation into existing synthetic toolkits. Proposals based on updated line arregression based models (as commonly seen in quantitative structure-property relationship or multi-linear regression modeling efforts) were not considered.
Funded projects will appear below as project overviews are posted to the website.
Advanced modeling capabilities will reduce developmental costs for new, environmentally friendly explosives, propellants, and pyrotechnics. Additionally, these models will reduce uncertainty for the potential environmental risks of novel materials and guide future toxicological studies to streamline data collection. A rapid assessment of materials will also help to select more environmentally sustainable energetic materials for further development.
The Department of Defense (DoD) is working to reduce the environmental and human health impacts of manufacturing, training, use and disposal of energetic materials. The DoD is also actively seeking alternative materials to eliminate known environmentally problematic chemicals such as RDX, perchlorate, lead (as a primary explosive or modifiers in propellants), and other potentially hazardous materials, while at the same time developing novel, more powerful and less sensitive materials to meet long-term mission goals. The most promising compounds rely on chemistries that are not commonly available and are not well characterized from an environmental, safety or occupational health standpoint. Such compounds include high nitrogen neutral compounds, metallic-organic complexes, or ionic compounds. Evaluation of developmental materials generally follows the phased approach for evaluation of energetics, as set forth in ASTM Guideline E2552-16, Standard Guide for Assessing the Environmental and Human Health Impacts of New Energetic Compounds. The first phase of this approach includes modeling to predict toxicological properties, and environmental fate and transport parameters.
Energetic molecules contain scaffolds and functional groups, and elemental ratios that are rarely found in pharmaceutical datasets commonly used in the public literature. Linear regression-based models have a long track record in cheminformatics, but open literature advances in recent years have shown the potential for advanced computational models to supplant traditional linear methods. This includes recent advances with deep neural networks, support vector machines, and gradient-boosted decision trees for toxicity predictions. These models are often highly nonlinear and do not pre-suppose a particular physical relationship between structure and property. They have also been shown to out-perform linearly parameterized QSPR models on a wide range of prediction problems.
The cost and time to meet the requirements of this SON are at the discretion of the proposer. Two options are available:
Standard Proposals: These proposals describe a complete research effort. The proposer should incorporate the appropriate time, schedule, and cost requirements to accomplish the scope of work proposed. SERDP projects normally run from two to five years in length and vary considerably in cost consistent with the scope of the effort. It is expected that most proposals will fall into this category.
Limited Scope Proposals: Proposers with innovative approaches to the SON that entail high technical risk or have minimal supporting data may submit a Limited Scope Proposal for funding up to $250,000 and approximately one year in duration. Such proposals may be eligible for follow-on funding if they result in a successful initial project. The objective of these proposals should be to acquire the data necessary to demonstrate proof-of-concept or reduction of risk that will lead to development of a future Standard Proposal. Proposers should submit Limited Scope Proposals in accordance with the SERDP Core Solicitation instructions and deadlines.