Peer Review of Draft NTP Approach to Genomic Dose-Response Modeling
Note: Individuals with disabilities who need accommodation must request it at least five business days prior to the meeting. TTY users should contact the Federal TTY Relay Service at 800-877-8339. Otherwise, contact Ms. Anna Stamatogiannakis for assistance.
Webinar 1: The NTP Proposed Approach to Genomic Dose-Response Modeling
August 30, 12:00–1:30 PM EDT
Speaker: Scott Auerbach, Ph.D., NIH/NIEHS/DNTP
NTP proposes to use an approach to genomic dose-response modeling that is consistent with the modeling approach recommended by the US EPA for continuous non-genomic endpoints and is implemented in the BMDS software. A three-step process is performed in which:
- Genomic features (genes/transcripts) are tested for dose-related response to test article treatment.
- Genomic features exhibiting a dose-related response are fitted to a variety of parametric models derived from the US EPA BMDS software. A best fit model for each genomic feature is then determined with methods consistent with those recommended by the US EPA. For each genomic feature, a benchmark dose (BMD) is determined based on a benchmark response threshold, specifically, a 10% increase or decrease in abundance of the genomic feature relative to controls. In addition, the BMD upper bound and lower bound are determined based on the best fit model for each feature.
- Genomic features and their associated BMD values are then placed into pre-defined gene sets (e.g., pathway gene sets) and a gene-set-level BMD is determined based on the mean and median BMD values for the features populating each gene set. In an identical manner, mean and median BMDU (upper bound of the BMD confidence interval) and BMDL (lower bound of the BMD confidence interval) for each gene set are determined.
NTP proposes to use the BMDExpress 2.0 software to carry out this analysis.
Webinar 2: Overview of the U.S. Army Approach to Genomic Dose-Response Modeling
September 1, 9:30-11:00 AM EDT
Speaker: Lyle Burgoon, Ph.D., U.S. Army
The US Army Engineer Research and Development Center has been investigating the use of toxicogenomics dose-response data to support material risk assessment for several years. This webinar will discuss the US Army’s latest approach, called the Good Risk Assessment Value for Environmental Exposures (GRAVEE). GRAVEE is based on the spline-based meta-regression technique approach for dose-response modeling, a nonparametric nonlinear modeling approach that is data-driven, rather than model-driven. GRAVEE is used to estimate points of departure from dose-response data. It is combined with Causal Adverse Outcome Pathway Networks (CAOPNs) to identify risk-relevant points of departure within a biological pathway context. Examples will be discussed for where this approach has been used with materials of interest to the military.
Webinar 3: Overview of the NC State Approach to Genomic Dose-Response Modeling
September 13, 3:30-5:00 PM EDT
Speaker: Fred Wright, Ph.D., N.C. State University
Analyses of high-throughput toxicogenomics data, in particular those using sequencing methods, require attention to read-calling algorithms and simplified dose-response modeling for datasets with relatively few samples. For cell lines treated with chemicals at varying concentrations, we describe a pipeline for handling expression data generated by TempO-Seq (as one possible technology) to align reads and clean/normalize raw count data. We follow with generic methods to identify differentially-expressed genes and calculate transcriptomic concentration-response points of departure. The methods are extensible to other forms of concentration-response gene expression data, and we discuss the utility of the methods for assessing variation in susceptibility and the diseased cellular state.
Webinar 4: An Automated Method to Identify Dose-Responsive Genes and Quantitate Points of Departure (PODs) from Transcriptomic Data
September 25, 2:00-3:30 PM EDT
Speaker: David Gerhold, Ph.D., NIH/NCATS
The National Center for Advancing Translational Sciences (NCATS) has developed an automated method to identify dose-responsive genes and quantitate points of departure (POD) from transcriptomic data. This method fits dose-response data to the Hill equation, to constrain the curve so it does not ‘over-fit’ to experimental noise. Replicates at each dose are tested for significant change from the vehicle controls, and for a significant slope of dose dependent increase or decrease in signal. The requirement for a significant slope minimizes false-positive calls at the lowest tested concentration. A minimum fold-change threshold is applied based on global variance to avoid false-positive calls that may arise from variable noise between genes. For genes that meet all three criteria, a POD is interpolated at the concentration where the curve deviates three SD from the vehicle controls. The Hill fit and three statistical tests comprise a conservative algorithm to minimize false-positive calls. For biphasic curves, the slope of the change at low concentration is tested first, given the importance of significant responses at low concentration. The identification of pathways that respond to a treatment evolves as annotation of pathways improves. These decisions can be made more conservative by requiring PODs for two genes in the same pathway. NCATS suggests that the POD algorithm be included in the BMD Express software package.
For questions or additional information, contact:
Anna Stamatogiannakis, Logistical Support
2635 Meridian Parkway
Durham NC, NC 27713
T: (919) 293-1652 (voice)
For questions or additional information contact:Mary S. Wolfe, Ph.D.
Deputy Division Director for Policy
Director, Office of Liaison, Policy, and Review
530 Davis Dr
Durham, NC 27713