Development and Improvement of ICE/ACE for Predictive Toxicology

Document Sample
Development and Improvement of ICE/ACE for Predictive Toxicology Powered By Docstoc
					Development and Improvement
         of ICE/ACE
  for Predictive Toxicology

           Sandy Raimondo
            Mace G. Barron
Office of Research and Development/NHEERL
            Gulf Ecology Division

            2 November 2005
             Overview
• Background: ICE/ACE development
• Current Research: ICE/ACE Validation
 and Improvement
ICE and ACE Software Development

Developed by Sonny Mayer (GED) and Colleagues


ICE (Interspecies Correlation Estimation)
  Estimates acute toxicity for a species, genus or
  family from a surrogate species


ACE (Acute to Chronic Estimation)
  Estimates chronic toxicity from raw acute
  toxicity data
                            ICE:
    Interspecies Correlation Estimations
Significance:
• Addresses second largest
source of variation in toxicity
data – variation of species
within a chemical

• Fills data gaps by estimating
toxicity of untested species
                How ICE works
Uses existing correlations of
toxicity values (LC50, LD50)
between a surrogate species
and a predicted taxa
(species, genus, or family)


 LC50: concentration that
       kills 50% of
       organisms

 LD50: dose that kills
       50%
                                   Acute toxicity estimates using
Cape Fear shiner (log 96-h LC50)     interspecies correlations


                                   3.0



                                   2.0

                                                           X2 = -0.025 + 1.098X1
                                   1.0                            r = 0.991




                                         1.0            2.0            3.0         4.0
                                               Rainbow Trout (log 96-h LC50)
ICE estimates LC50 from surrogate species
LC50 and available species correlation
 Current ICE Uses: EPA Program Offices

• Office of Water (OW): draft Ambient Water
  Quality Criteria (AWQC) guidelines,
  endangered species

• Office of Pesticide Programs (OPP):
  qualitative use in risk assessment currently
  being implemented.
          ICE: Applications
Ecological Risk Assessment (ERA)
  • Generate species sensitivity distribution
  • Define risk management level


Endangered Species Risk Assessment
  • Surrogate test species are toxicologically
    representative of endangered species (Mayer et al.)
  • ICE can be used to estimate toxicity to T&E species
    using existing correlations (147 LC50s; 20 species)
                            ACE:
           Acute to Chronic Estimations

Significance:
Provides estimated chronic
toxicity for species with only
acute data


Acute: ie.96-hour LC50/ LD50

Chronic: long-term, sublethal
                            ACE:
          Acute to Chronic Estimations

Application:
• Reduced reliance on
acute to chronic ratios
based on multiple species
and chemicals

• Being considered by OPP
for qualitative use in risk
assessment
How ACE Works
ACE Chronic Mortality Prediction
Uses raw survival data to estimate chronic mortality

            Linear Regression                                 Accelerated Life
              Analysis (LRA)                                   Testing (ALT)

                      10                                       10
at each time period




                                                                8
 Log NOEC (µg/L)




                      8.0                                EXTRA  6
                                                       RESPONSE
                                                           (%)  4
                      6.0                                       2
                                                                 0
                      4.0                                     0.140
                                    a=NOEC                         12.0
                      2.0                                              1.0
                                                                         0.08                                          100
                                                                   DOSE 0.06                                      80
                                                                   (ug/L)                                    60
                            0.010     0.021    0.042                          0.04                   40
                                                                                 0.02           20        TIME
                                    1/time (h)                                          0.0 0             (DAY)
   Limitations of ICE and ACE

• validation and uncertainty (ICE, ACE)

• data poor (e.g., ICE wildlife)

• flexibility-power (ICE)
      Future Direction of ICE/ACE
                     Expand Datasets
                        (QA/QC)


     ICE                                        ACE

     Validation,
Refinement, Expansion                        Validation



      New Software


               Improved Tool Functionality
   ICE Improvement Procedure
1. Model Validation: Regression analysis
2. Model Refinement: Stepwise regression
3. Model Expansion: Power and number of
   models
4. QA/QC: Bootstrap validation
5. Defining Assumptions / Developing user
   guidelines: multivariate analyses
6. New Software
   ICE Improvement Procedure
1. Model Validation:
   Assessment of existing significant correlations
        Aquatic ICE models (P<0.01)
            •   Species: 565
            •   Genus: 195
            •   Family: 291

  Validation dataset for aquatics
        Ambient Water Quality Criteria (AWQC) data
            •   88 chemicals (12 pesticides)
            •   279 species
            •   1458 new data points
                         ICE Validation Example
                                   Rainbow Trout Surrogate - Pesticides only


                          6

                                          2
                          5              R = 0.95
Log10 Actual LC50




                          4

                          3

                          2                                                        < 3x difference
                                                                                   > 3x difference
                          1

                          0
                    -1         0              1         2            3         4   5
                          -1

                          -2
                                        Log10 ICE predicted LC50
     ICE Improvement Procedure
2. Model Refinement                                                   Rainbow Trout Surrogate - Pesticides only


                                                             6
   Stepwise Regression                                       5
                                                                             2
                                                                            R = 0.95




                                    Log10 Actual LC50
     •   Finds the best fit model                            4

     •   Quality of data                                     3


     •   Power of model                                      2


     •   Mode of action                                      1

                                                             0
                                                        -1        0              1         2            3         4   5
                                                             -1

                                                             -2
                                                                           Log10 ICE predicted LC50



 Model improvement through reduction (filter by MOA,
 species, data quality)
     ICE Improvement Procedure
3. Model Expansion
   Increase power of the models
     • Many existing models have N=3
     • More data increases likelihood of significant correlations

  Increase the number of significant correlations
               ICE Wildlife Correlations (P ≤ 0.01)
                       Original   Current
                         62         152


 Model improvement through addition (species, chemical)
    ICE Improvement Procedure
4. QA/QC of all refined and improved models
  Bootstrap validation of all improved models
  • data points are                                           Rainbow Trout Surrogate - Pesticides only

    randomly removed
                                                     6
  • model is recreated                                               2
                                                     5              R = 0.95

  • removed data are used
                            Log10 Actual LC50
                                                     4
    to validate model
                                                     3
  • up to 1000 replicates
                                                     2

                                                     1

                                                     0
                                                -1        0              1         2            3         4   5
                                                     -1

                                                     -2
                                                                   Log10 ICE predicted LC50
        ICE Improvement Procedure
                                                               Rainbow Trout Surrogate - Pesticides only
Multivariate Analysis
   •   mode of action                                 6
                                                                                     2
                                                                                   R = 0.95
   •   chemical class                                 5




                             Log10 Actual LC50
   •   species                                        4

   •   life history                                   3

                                                      2

   • model fit (yes or no)                            1

   • degrees of freedom                               0
                                                 -1        0            1           2            3         4   5
   • model R2, p-value                                -1

                                                      -2
                                                                        Log10 ICE predicted LC50



                        User Guidance
  ICE Improvement Procedure
6. Updated Software (2007)
  Selectable datasets
  Interactive program
       user can enter in new dataset for species not in
       ICE (ie. endangered species) and ICE will build
       predictive model based on internal data

  Broader applicability
  Wider user base
  ACE Validation and User Guidance
Validate using expanded dataset
   • Original validation used 30 fish acute: chronic data pairs
   • Expanded validation dataset
       >150 datasets
                                                                       5
Define where models are robust                                         4
                                                                               R2 = 0.95


   • MOA



                                      Log10 Survival MATC
                                                                       3
   • Chemical Classes
   • Species                                                           2
                                                                                               < 5x difference
                                                                       1                       > 5x difference

                                                                       0

Develop new user guidance                                   -4   -2
                                                                      -1
                                                                           0          2    4          6


   • Multivariate Analyses
                                                                      -2
   • Improve value to user                                            Log10 ACE ALT
           2006 Anticipated
        Products & Outcomes:
•   Manuscript: Wildlife Toxicity Estimation
•   Manuscript: Validation of ICE
•   Manuscript: Validation of ACE
•   Technology transfer

2007: Updated ICE Software

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:11/23/2011
language:English
pages:25