Imprecision and importance: Probabilistic graphical models in toxicology
Lead Research Organisation:
University of York
Department Name: Mathematics
Abstract
To evaluate the safety of a new chemical existing toxicological protocols typically require at least 4 years of research, £3-4 million in funding, and experiments on over 4000 animals. This is inefficient, ethically questionable, and fails to exploit recent biological, mathematical and computational advances. We plan to use advanced computational and statistical methods to investigate how the evaluation process can be improved.
More explicitly, we will use Bayesian networks to exploit existing data both to identify the key studies necessary for efficient toxicological assessment, and to quantify what levels of imprecision should be tolerated. Bayesian methods allow us to take existing knowledge about toxicity, even if this knowledge is incomplete or inaccurate, and use it to make better predictions for future chemicals. Because uncertainty and imprecision can be built very naturally into the models (indeed, they are necessary), it becomes easier to make data-driven probabilistic statements about toxicological risk.
The outputs will provide a rigorous quantification of the value of each element of existing protocols. In doing so, we will seek to answer the question of how far the detailed data from a panel of 50+ well-studied toxicants can justify applying the 3Rs (the reduction, refinement and replacement of animal experiments) in the toxicological testing necessary for future chemicals. Explicitly, reduction of the number of animals used for testing will be achieved where our models indicate that enough precision can be derived from a smaller battery of tests; refinement of animal testing protocols will become possible where our models idenitify efficiencies through the holistic assimilation of broad-spectrum data; and where appropriate replacement of animal tests by can be rationally and quantifiably justified early in a given chemical's testing strategy.
More explicitly, we will use Bayesian networks to exploit existing data both to identify the key studies necessary for efficient toxicological assessment, and to quantify what levels of imprecision should be tolerated. Bayesian methods allow us to take existing knowledge about toxicity, even if this knowledge is incomplete or inaccurate, and use it to make better predictions for future chemicals. Because uncertainty and imprecision can be built very naturally into the models (indeed, they are necessary), it becomes easier to make data-driven probabilistic statements about toxicological risk.
The outputs will provide a rigorous quantification of the value of each element of existing protocols. In doing so, we will seek to answer the question of how far the detailed data from a panel of 50+ well-studied toxicants can justify applying the 3Rs (the reduction, refinement and replacement of animal experiments) in the toxicological testing necessary for future chemicals. Explicitly, reduction of the number of animals used for testing will be achieved where our models indicate that enough precision can be derived from a smaller battery of tests; refinement of animal testing protocols will become possible where our models idenitify efficiencies through the holistic assimilation of broad-spectrum data; and where appropriate replacement of animal tests by can be rationally and quantifiably justified early in a given chemical's testing strategy.
Technical Summary
Three related work packages, WP1-3, which are interdependent and overlapping.
Approximate timings WP1: Jan - Dec 2013; WP2: Jan - Sep 2014; WP3: Sep 2014 - Sep 2015.
WP1: Practical PGMs for toxicology
- Familiarisation with existing methods of toxicological terminology, assessment schemes and datasets
- Training in Bayesian statistics, machine learning, network theory, extreme value theory, coding for PGM implementation.
- Computational implementation and validation of simple strategic PGMs on toxicology.
- Extensions: alternative learning algorithms, model averaging, sequential decision theory.
- Proof-of-concept study for Repeated Dose Toxicity.
WP2: Theoretical frameworks
- Adding distributional information to simple probabilistic models: Theoretical considerations of uncertainty and precision in simple networks, rules-of-thumb for adaptive testing strategies.
- Complex systems, simple mechanisms? Exploration of the possibility of simple stochastically parameterised mechanistic models to explain broad scale toxicological phenomena.
WP3: Building an integrated Bayesian framework for toxicology testing
Synthesis of WP1 and 2 results, together with inputs and refinements from industrial contacts and collaborators, so as to arrive at practical solutions for the reliable replacement, reduction and refinement of animal experiments in toxicology based on a wider range of toxicological outcomes: toxico-kinetics, skin sensitisation, carcinogenicity, and reproductive toxicity. We anticipate that the resulting integrated testing strategies will be adaptive; identifying conditional independence within the data has the potential for reliable and safe reduction, replacement and refinement of animal testing.
Outputs: Research papers in high ranking specialist toxicology and more general interdisciplinary journals; dissemination through conferences, industrial contacts and collaborative research visits, and dedicated modelling workshop in 2014.
Approximate timings WP1: Jan - Dec 2013; WP2: Jan - Sep 2014; WP3: Sep 2014 - Sep 2015.
WP1: Practical PGMs for toxicology
- Familiarisation with existing methods of toxicological terminology, assessment schemes and datasets
- Training in Bayesian statistics, machine learning, network theory, extreme value theory, coding for PGM implementation.
- Computational implementation and validation of simple strategic PGMs on toxicology.
- Extensions: alternative learning algorithms, model averaging, sequential decision theory.
- Proof-of-concept study for Repeated Dose Toxicity.
WP2: Theoretical frameworks
- Adding distributional information to simple probabilistic models: Theoretical considerations of uncertainty and precision in simple networks, rules-of-thumb for adaptive testing strategies.
- Complex systems, simple mechanisms? Exploration of the possibility of simple stochastically parameterised mechanistic models to explain broad scale toxicological phenomena.
WP3: Building an integrated Bayesian framework for toxicology testing
Synthesis of WP1 and 2 results, together with inputs and refinements from industrial contacts and collaborators, so as to arrive at practical solutions for the reliable replacement, reduction and refinement of animal experiments in toxicology based on a wider range of toxicological outcomes: toxico-kinetics, skin sensitisation, carcinogenicity, and reproductive toxicity. We anticipate that the resulting integrated testing strategies will be adaptive; identifying conditional independence within the data has the potential for reliable and safe reduction, replacement and refinement of animal testing.
Outputs: Research papers in high ranking specialist toxicology and more general interdisciplinary journals; dissemination through conferences, industrial contacts and collaborative research visits, and dedicated modelling workshop in 2014.
Planned Impact
Impact summary
While this research presents unashamedly interesting mathematical, statistical and computational challenges, we are committed to carrying out research of genuine and lasting value to the toxicological community, and to the NC3Rs agenda in particular.
Our commitment to active collaboration with industry is evidenced by two letters of support (attached), from Dr Dick Lewis (Global Head of Toxicology and Health Science, Syngenta) and Dr Joanna Jaworska (Modeling & Simulation, Procter and Gamble). We will use these contacts as a basis to explore further industrial collaboration, as well as planning to collaborate on some of the technical aspects raised in Dr Jaworska's recent publications. These links have the potential to develop into lasting research relationships with high impact (and potentially high financial value in terms of 3Rs-related savings).
We plan an intensive hands-on workshop toward the end of 2014, to which circa 10 delegates will be invited from across industry and regulatory bodies. The aim of this will be to critically assess the results generated, and to steer the final phase of the project toward useful outputs. We will initiate feedback and follow up collaborations where possible, with a view both to extracting the maximum value from this research and to initiating subsequent research programmes.
We plan to publish our findings in a variety of academic journals, including those of direct relevance to the toxicology community. In addition, where appropriate we will seek involvement in more industry-led policy papers. We plan to present our overall findings at the International Congress on Toxicology in 2015, and will also present work-in-progress reports at smaller conferences in toxicology, and in more general mathematical biology, in the UK and Europe (note that the MMEE conference is in York in 2013, organised by JWP; the appointed PDRA will be fully involved).
Overall we anticipate that industry will receive a significant financial benefit as a result of this research programme. As previously noted, the typical cost of a safety evaluation for a new chemical is £3-4M. Very large savings can potentially be made (relative to the cost of this project) even by the reduction or refinement of a subset of animal tests for particular cases within an adaptive testing framework (as noted in the letter of support from Dick Lewis, Syngenta). This financial benefit is in addition to the important moral and ethical benefits involved with the general reduction of reliance on animal testing.
The University of York encourages public engagement and outreach work, and funding is available tfor support and training where necessary. JWP and members of his research group have presented their research findings in a variety of locations including schools, science festivals, and pubs, and we will continue to foster and exploit such opportunities.
While this research presents unashamedly interesting mathematical, statistical and computational challenges, we are committed to carrying out research of genuine and lasting value to the toxicological community, and to the NC3Rs agenda in particular.
Our commitment to active collaboration with industry is evidenced by two letters of support (attached), from Dr Dick Lewis (Global Head of Toxicology and Health Science, Syngenta) and Dr Joanna Jaworska (Modeling & Simulation, Procter and Gamble). We will use these contacts as a basis to explore further industrial collaboration, as well as planning to collaborate on some of the technical aspects raised in Dr Jaworska's recent publications. These links have the potential to develop into lasting research relationships with high impact (and potentially high financial value in terms of 3Rs-related savings).
We plan an intensive hands-on workshop toward the end of 2014, to which circa 10 delegates will be invited from across industry and regulatory bodies. The aim of this will be to critically assess the results generated, and to steer the final phase of the project toward useful outputs. We will initiate feedback and follow up collaborations where possible, with a view both to extracting the maximum value from this research and to initiating subsequent research programmes.
We plan to publish our findings in a variety of academic journals, including those of direct relevance to the toxicology community. In addition, where appropriate we will seek involvement in more industry-led policy papers. We plan to present our overall findings at the International Congress on Toxicology in 2015, and will also present work-in-progress reports at smaller conferences in toxicology, and in more general mathematical biology, in the UK and Europe (note that the MMEE conference is in York in 2013, organised by JWP; the appointed PDRA will be fully involved).
Overall we anticipate that industry will receive a significant financial benefit as a result of this research programme. As previously noted, the typical cost of a safety evaluation for a new chemical is £3-4M. Very large savings can potentially be made (relative to the cost of this project) even by the reduction or refinement of a subset of animal tests for particular cases within an adaptive testing framework (as noted in the letter of support from Dick Lewis, Syngenta). This financial benefit is in addition to the important moral and ethical benefits involved with the general reduction of reliance on animal testing.
The University of York encourages public engagement and outreach work, and funding is available tfor support and training where necessary. JWP and members of his research group have presented their research findings in a variety of locations including schools, science festivals, and pubs, and we will continue to foster and exploit such opportunities.