Division of Biometry and Risk Assessment

1
Division of Biometry and Risk Assessment

• John Appleget Computer Specialist
• James Chen, Ph.D. Mathematical Statistician
• Yi-Ju Chen Post Doc
• Robert Delongchamp, Ph.D. Mathematical
  Statistician
• Ralph Kodell, Ph.D. Director
• Daniel Molefe, Ph.D. Post Doc
• Bruce Pearce Computer Specialist
• Susan Taylor Program Support Specialist
• Angelo Turturro, Ph.D. Research Biologist
• Cruz Velasco, Ph.D. Post Doc
• John Young, Ph.D. Research Biologist
• Qi Zheng, Ph.D. Staff Fellow

2
Research Highlights

• Fumonisin B1 Risk Modeling
• Cryptosporidium parvum Study
• Cumulative Risk for Chemical Mixtures
• Computational Toxicology
• Photocarcinogenicity Theory Methods
• Analysis of cDNA Microarray Data
• Staff Enrichment

3
Fumonisin B1 Risk Modeling
Qi Zheng et al.

• NTP IAG Study in rats and mice (P. Howard)
• Liver tumors in female micekidney tumors in male
  rats
• Directed/encouraged by Bern Schwetz
• CFSAN, CVM
• Two recommendations of SAB SVT
• Project related to Food Safety Initiative
• Project for intra-division collaboration

4
Female Mouse Liver Tumors

• Adjusted tumor rates at 104 weeks
• Hepatocellular adenoma or carcinoma

Probability
ppm
5
Mathematical Model

• Use MVK two-stage, cell-proliferation model to
  predict probability of tumor at 104 weeks

?(t)
?1
?2
Normal N(t)
Malignant
Preneoplastic
?(t)
6
Hypothesis

• Fumonisin B1 affects the incidence ofliver tumor
  formation in mice byincreasing the death rate of
  cellswhich leads tocompensatory proliferation.

7
Implementing the Model

• Use allometric relationship between liver weight
  and body weight, LW(t)aBW(t)b,
  to estimate the liver weight
• Estimate the number of cells in the liver by
  N(t)LW(t)/CW
• Estimate the net growth rate of the liver using
  dlogLW(t)/dt

8
Implementing the Model

• Use PCNA data to estimate the cell birth
  rate, ?(t)
• Estimate the cell death rate by
  ?(t)?(t)-dlogLW(t)/dt

9
Implementing the Model

• Relate differential effect of FB1 on ?(t), and,
  consequently, ?(t) by level of sphinganine in
  liver
• Infer mutation rates, ?1 and ?2, (constant w.r.t.
  FB1 and time) from tumor data

10
Female Mouse Liver Tumors

• Tumor incidence at 104 weeks
• Hepatocellular adenoma or carcinoma
• Observed .117, .065, .021, .427, .883
• Predicted .091, .084, .105, .284, .992

Probability
ppm
11
Male and Female Mouse Liver Tumors
Male
Observed .268, .211, .190, .213, .213 Predicted
.199, .201, .198, .233, .237 Observed .117,
.065, .021, .427, .883 Predicted .091, .084,
.105, .284, .992
Female
Probability
ppm
12
Fumonisin B1 Summary

• Data and model are consistent with hypothesis
• FDA Workshop on Fumonisins Risk Assessment
  February, 2000
• Food Additives and Contaminants, 2001
• FAO/WHO JECFA (Feb., 2001) used extensively in
  draft report on fumonisins CFSAN (Mike Bolger)
• Model kidney tumor risk in male rats?

13
Cryptosporidium parvum Study
Angelo Turturro et al. E07082.01

• IAG with EPA-NCEA, Cincinnati - B. Boutin
• Much input from CFSAN (R. Buchanan, G. Jackson,
  M. Miliotis)
• New challenge for NCTR
• Cryptosporidium parvum is a protozoan
• Common contaminant of drinking water
• Can also contaminate the food supply

14
Objectives

• To develop a model for transmission dynamics of
  Cryptosporidium parvum in human outbreaks
• To standardize the dose of Cp strains in the
  neonatal mouse (three isolates)
• To establish an appropriate animal model
• Brown Norway rat
• Chemically supressed C57Bl/6 mouse (Dex)

15
Objectives (cont.)

• To investigate subpopulations with varying
  degrees of immunocompetence
• Three age groups - young, adult, elderly
• Pregnant
• Immunosuppressed similar to AIDS
• Physiologically stressed - diet, exercise
• Status Protocol reviewed, revised, re-submitted

16
Cumulative Risk for Chemical Mixtures
James Chen, Yi-Ju Chen et al. E07087.01

• IAG with EPA-NCEA, Cincinnati- G. Rice, L.
  Teuschler
• Objective To develop and apply a Relative
  Potency Factor (RPF) methodology for estimating
  the cumulative risk from exposure to a mixture of
  chemicals having a common mode of action (e.g.,
  organophosphates cholinesterase inhibition)
  FQPA, 1996

17
Specific Aims

• To use an expanded definition of dose addition to
  develop a risk estimation method that does not
  depend strictly on parallelism of
  log-dose-response curves
• To develop a classification algorithm for
  clustering chemicals into several constant
  relative potency subsets

18
Advantages

• Uses actual dose-response functions of mixture
  components, not just ED10s, say (like TEF, HI,
  etc.)
• If the RPF is constant across all chemicals, then
  invariant to choice of index chemical
• Can be used even when the RPF differs for
  different subsets of chemicals in the mixture
• Status Protocol in review

19
Computational Toxicology
John Young et al. E07083.01

• Objective To develop an expert computational
  system for prediction of organ-specific rodent
  carcinogenicity by applying structure activity
  relationships (SAR) in conjunction with data on
  short-term toxicity tests (STT) and nuclear
  magnetic resonance (13C-NMR) spectroscopy.

20
Motivation

• FDAs need to
• bring safe products to market more quickly
• screen out unsafe products reliably
• CFSAN (M. Cheeseman)
• streamline toxicity testing, e.g., require
  sponsor to conduct target-specific toxicity based
  on systems prediction

21
Database

• 1298 chemicals in Carcinogenic Potency Database
• Group 1 carcinogenicity in liver
• Group 2 carcinogenicity, but not in liver
• Group 3 no carcinogenicity in any organ
• Add data on SAR, STT and NMR

22
Database (cont.)

• 392 NTP chemicals in CPDB
• 342 positive in liver for ? 1 species-sex combo.
• For good mix of positive/negative, might need to
  do
• species-specific prediction
• sex-specific prediction

23
Strategy

• Training set
• Use 392 NTP chemicals in CPDB
• Testing set
• Use 288 literature chemicals in CPDB
• Use 282 pharmaceuticals in CDER database
• 33 positive in liver for ? 1 species-sex combo.
• Status Protocol recently approved and
  implemented

24
Photocarcinogenicity Theory Methods
Ralph Kodell, Daniel Molefe et al. E07061.01

• FDA
• CFSAN Cosmetics
• CDER Drugs (K. Lin)
• NCTRs Phototoxicity Program (P. Howard)
• CRADA w/ ARGUS Laboratory S00213
• Post Doc funding through NTP E02037.01

25
Statistical Approaches

• Standard Testing Method
• Logrank test for differences in distributions of
  time to first observed tumor
• New Testing Method
• Test for difference in number of induced tumors
• Test for difference in distributions of time to
  observation of tumors

26
Accomplishments/Plans

• Model developed for repeated-exposure case
• Computational optimization procedure developed
• Data on first of eight Argus studies analyzed
• Compare to logrank and Dunsons method
• Status Ongoing.

27
Analysis of cDNA Microarray Data
Bob Delongchamp, Cruz Velasco et al. E07096.01

• cDNA Microarrays
• popular new biotech tool
• vast amounts of data on gene expression quickly
• Statistical issues
• Experimental design
• Analysis and interpretation

28
Statistical Issues

• Experimental design
• Replication arrays and genes
• Data analysis
• Adjustment for nuisance sources of variation
• Appropriate methods for assessing differences
• Adjustment for multiple comparisons
• Identification of genetic profiles

29
Figure 1. Intensities observed in rat
hepatocytes. Upper Right - Untreated
Array Lower Left - MP Treated Array Lower
Right - PM Treated Array
30
Figure 2. Array maps of log(Iga/Ig). Upper
Right - Untreated Array Lower Left - MP
Treated Array Lower Right - PM Treated Array
31
Figure 3. Intensities adjusted within 6x6
blocks. Upper Right - Untreated Array Lower
Left - MP Treated Array Lower Right - PM
Treated Array
32
Figure 4. Intensities adjusted for
splotches (Ka) and saturation (Ka). Upper
Right - Untreated Array Lower Left - MP
Treated Array Lower Right - PM Treated Array
33
Objectives

• Data analysis
• Appropriate methods for assessing differences
• Individual genes
• Clusters of genes (profiles)
• Adjustment for multiple comparisons
• PCER, FWER, FDR
• Status Protocol in development

34
Staff Enrichment

• Short courses and conferences
• UCLA Functional Genomics (Chen)
• IBS/ENAR Conference (Chen, Delongchamp, Kodell)
• Gordon Conference on Bioinformatics (Zheng)
• Genetic and Evolutionary Computation Conference
  (Pearce)
• IAG with UAMS (R. Evans)

35
Staff Enrichment

• Lab visits
• Academia Sinica, Taiwan (Chen, 2 weeks)
• Visualization, classification (C-H Chen)
• Jackson Lab. (Delongchamp, 1 month)
• Differential gene expression (G Churchill)
• Visits to other FDA Centers
• CDRH (Greg Campbell) Delongchamp, Velasco,
  Harris
• Visiting scientists