1. Introduction to Biostatistics and Bioinformatics Experimental Design

2. Experimental Design by Christine Ambrosino www.hawaii.edu/fishlab/Nearside.htm

3. Experimental Design Overcoming the threat from chance and bias to the validity of conclusion.

4. Experimental Design Inputs Process Outputs Controllable Factors Uncontrollable Factors

5. Experimental Design Recognition and statement of the problem (e.g. testing a specific hypothesis or open ended discovery). Selecting a response variable. Choosing controllable factors and their range. Listing uncontrollable factors and estimate their effect. Choosing experimental design. Performing experiment. Statistical analysis of data. Designing the next experiment based on the results.

6. Exploring the Parameter Space One factor at a time Factor 1 Score Factor 2 Score Factor 3 Score Factor 1 Factor 2 2-factor factorial design3-factor factorial design k-factor factorial design (2 k experiments) k factors : 2k experiments 4 experiments 8 experiments For example, 7 factors: 128 experiments, 10 factors: 1,024 experiments

7. Randomization Statistical methods require that observations are independently distributed random variables. Randomization usually makes this assumption valid. Randomization guards against unknown and uncontrolled factors. Randomize with respect to analysis order, location, material etc. Order of Measurements p = 0.19p = 0.32 Not Randomized Randomized No change in sensitivity during measurement

8. Randomization Order of Measurements p = 0.19p = 0.32 Not Randomized Randomized Order of Measurements p = 5.7x10 -6 No change in sensitivity during measurement Change in sensitivity during measurement p = 0.20 Standard Deviation: 0.8, 0.8 Standard Deviation: 0.7, 0.9 Standard Deviation: 1.8, 1.3

9. Blocking Blocking is used to control for known and controllable factors. Randomized Complete Block Design - minimizing the effect of variability associated with e.g. location, operator, plant, batch, time. The Latin Square Design - minimizing the effect of variability associated with two independent factors The rows and columns represent two restrictions on randomization

10. Replication Replication is needed to estimate the variance in the measurements. Technical replicates (repeat measurements). Process replicates Biological replicates

11. Uncertainty in Determining the Mean ComplexNormalSkewedLong tails n=3 n=10 Mean n=100 n=3 n=10 n=100 n=3 n=10 n=100 n=10 n=100 n=1000

12. An example of bad experimental design

13. Protein Identification and Quantitation by Mass Spectrometry Mass Spectrometry m/z intensity Identity Quantity Samples Peptides

14. A proteomics example – no replicates three replicates Log 2 Sum Spectrum Count Log 2 Spectrum Count Ratio

15. Analytical Measuments: Precision and Accuracy Theoretical Concentration Measured Concentration

16. Testing multiple hypothesis Is the concentration of calcium/calmodulin-dependent protein kinase type II different between the two samples? What protein concentration are different between the two samples? p = 2x10 -6 The p-value needs to be corrected taking into account the we perform many tests. Bonferroni correction: multiply the p-value with The number of tests performed (n): p corr = p uncorr x n In this case where 3685 proteins are identified, so the Bonferroni corrected p-value for calcium/calmodulin-dependent protein kinase type II is p corr = 2x10 -6 x 3685 = 0.007

17. Testing multiple hypothesis The p-value distribution is uniform when testing differences between samples from the same distribution. Normal distribution Sample size = 10 p-value 1 0 # of test p-value 1 0 # of test p-value 1 0 # of test 0 8 0 60 0 500 10,000 tests1,000 tests100 tests

18. Testing multiple hypothesis The p-value distribution is uniform when testing differences between samples from the same distribution. Normal distribution Sample size = 10 30 tests from a distribution with a different mean (μ 1 -μ 2 >>σ) p-value 1 # of test p-value 1 # of test p-value 1 0 # of test 0 30 0 100 0 500 10,000 tests1,000 tests100 tests 0 0

19. Testing multiple hypothesis Controlling for False Discovery Rate (FDR) Normal distribution Sample size = 10 30 tests from a distribution with a different mean (μ 1 -μ 2 >>σ) p-value 1 False Rate p-value 1 False Rate p-value 1 0 False Rate 0 1 0 1 0 1 0 0 False Discovery Rate False Discovery Rate False Discovery Rate 10,000 tests1,000 tests100 tests

20. Testing multiple hypothesis False Discovery Rate (FDR) and False Negative Rate (FNR) Normal distribution Sample size = 10 100 tests 30 tests from a distribution with a different mean p-value 1 False Rate p-value 1 False Rate p-value 1 0 False Rate 0 1 0 1 0 1 0 0 μ 1 -μ 2 =2σμ1-μ2=σμ1-μ2=σμ 1 -μ 2 =σ/2 False Discovery Rate False Negative Rate False Discovery Rate False Negative Rate False Discovery Rate False Negative Rate

21. Sampling – Gaussian Peak Retention Time Intensity

22. Sampling – Gaussian Peak

23. Definition of a molecular signature FDA calls them “in vitro diagnostic multivariate assays” A molecular signature is a computational or mathematical model that links high-dimensional molecular information to phenotype or other response variable of interest.

24. 1.Models of disease phenotype/clinical outcome Diagnosis Prognosis, long-term disease management Personalized treatment (drug selection, titration) 2.Biomarkers for diagnosis, or outcome prediction Make the above tasks resource efficient, and easy to use in clinical practice 3.Discovery of structure & mechanisms (regulatory/interaction networks, pathways, sub- types) Leads for potential new drug candidates Uses of molecular signatures

25. Oncotype DX Breast Cancer Assay Developed by Genomic Health (www.genomichealth.com) 21-gene signature to predict whether a woman with localized, ER+ breast cancer is at risk of relapse Independently validated in thousands of patients So far performed >100,000 tests Price of the test is $4,175 Not FDA approved but covered by most insurances including Medicare Its sales in 2010 reached $170M and with a compound annual growth rate is projected to hit $300M by 2015.

26. EF Petricoin III, AM Ardekani, BA Hitt, PJ Levine, VA Fusaro, SM Steinberg, GB Mills, C Simone, DA Fishman, EC Kohn, LA Liotta, "Use of proteomic patterns in serum to identify ovarian cancer", Lancet 359 (2002) 572–77

27. Check E., Proteomics and cancer: running before we can walk? Nature. 2004 Jun 3;429(6991):496-7.

28. Example: OvaCheck Developed by Correlogic (www.correlogic.com) Blood test for the early detection of epithelial ovarian cancer Failed to obtain FDA approval Looks for subtle changes in patterns among the tens of thousands of proteins, protein fragments and metabolites in the blood Signature developed by genetic algorithm Significant artifacts in data collection & analysis questioned validity of the signature: -Results are not reproducible -Data collected differently for different groups of patients http://www.nature.com/nature/journal/v429/n6991/full/42 9496a.html

29. Main ingredients for developing a molecular signature

30. Base-Line Characteristics DF Ransohoff, "Bias as a threat to the validity of cancer molecular-marker research", Nat Rev Cancer 5 (2005) 142-9.

31. How to Address Bias DF Ransohoff, "Bias as a threat to the validity of cancer molecular-marker research", Nat Rev Cancer 5 (2005) 142-9.

32. Experimental Design - Summary Chance and bias is a threat to the conclusions from experiments Controllable and uncontrollable factors Randomization to guard against unknown and uncontrolled factors Replication (technical, process, and biological replicates) is used to estimate error in measurement and yields a more precise estimate. Blocking to control for known and controllable factors Multiple testing Molecular markers

33. Experimental Design - Summary Use your domain knowledge: using a designed experiment is not a substitute for thinking about the problem. Keep the design and analysis as simple as possible. Recognize the difference between practical and statistical significance. Design iterative experiments.