Productive long term collaborations can be built on our powerful - PowerPoint PPT Presentation

Productive long ‐ term collaborations can be built on our powerful statistical support Qin Liu MD, Ph.D. The Wistar Institute Statistical Practice in Cancer Conference March 1st, 2019

Statistician’s task  Provide critical collaborative statistical support for diverse biomedical projects – Grant applications – Research designs – Data analysis results interpretations – Manuscripts  Develop novel statistical approaches for accurate data analysis

Example 1 Analysis of Mouse Tumor Growth Data Determine treatment effects using longitudinal tumor growth data from a mouse model

Tumor Growth Data from A Mouse Model Experimental Procedure Randomization 10 mice / treatment Treatment Arm 1 Treatment Arm 2 Measure tumor volume over time

Example: Problems with Longitudinal Tumor Growth Data Analysis 10 mice / treatment Control Single (GW) Single (PLX) Combo ANOVA compares fold-change Problem: The ANOVA test outcome at a single time point. It does not detect a statistically significant difference in longitudinal tumor growth between the BRAF inhibitor (PLX4720) alone vs. in combination with the CSF1R inhibitor (GW2580)

The ANOVA test compares fold ‐ changes of tumor volume from baseline at 14 days One-way ANOVA with Bonferroni’s adjusted p-values were reported Control GW2580 PLX4720 GW2580 0.911 PLX4720 0.007 0.277 GW2580 + PLX4720 <0.001 0.005 0.693

Mouse Tumor Growth Data 10 mice / treatment Control Single (GW) Single (PLX) Combo Compare fold-change Problem: The ANOVA test compares a single time point. It does not detect a statistically significant difference in longitudinal tumor growth between the BRAF inhibitor (PLX4720) alone vs. in combination with the CSF1R inhibitor (GW2580)

Solution: Compare Tumor Growth Trends (Velocities) Using A Fitted Mixed model 800 700 Bonferroni’s 600 Tumor volume adjusted 500 Control p=0.002 400 GW p<0.001 300 PLX 200 p=0.012 Combination 100 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 days Predicted trend (Control) Observed mean (Control) Predicted trend (GW2580) Observed mean (GW2580) Predicted trend (PLX4720) Observed mean (PLX4720) Predicted trend (Combined) Observed mean (Combined) Wang,…, Liu, …Herlyn, Kaufman. BRAF Inhibition Stimulates Melanoma-Associated Macrophages to Drive Tumor Growth. Clinical Cancer Research . 2015.

Example 2 Clinical Data Analysis Examine the effect of aging on the response of melanoma to therapy

Results From A Pre ‐ clinical Study ‐ Mouse Model of Melanoma • Tumors in young mice grew faster but they responded robustly to BRAF inhibitor PLX4720 • Tumors in aged mice responded poorly to BRAF inhibitor (PLX)

Clinical Data Analysis Data mining with stratifying patients by treatment and treatment history In patients treated by Vemurafenib without prior therapies, older individuals show muted tumor response to BRAF inhibitor Vemurafenib. Kaur, …, Liu, …, Weeraratna. sFRP2 in the aged microenvironment drives melanoma metastasis and therapy resistance. Nature 2016; 532: 250-266.

Example 3 High Throughput Screening Data Analysis How can we evaluate the potential synergistic effect of two drugs from in vitro data efficiently and accurately? A new statistical approach development and applications

Why Do We Study Drug Combinations? Limitations of single drug treatment • A single drug often has limited anti ‐ tumor effect • Resistance is a major issue Advantages of treatment with drug combination • May achieve a desired efficacy at lower dose with less side effects • May reduce the resistance • May have Synergistic effect

Drug Combinations Can Target the Same or Different Biological Pathways D C A Loewe additivity B Bliss independence Modified from Chudnovsky, Y. et al. Journal of Clinical Investigation. 2005;115:813-824

Definition of Synergism Bliss independence criterion • Additive effect – For given doses, the combined response (inhibition rate of cancer cell growth) from two drug compounds equals the sum of each single drug response subtracted by the multiplication of each single drug response. � 1,2 = � 1 + � 2 – � 1 * � 2 , 0< � <1 • Synergistic effect – the observed response from combination of two compounds exceeds their additive effect (excess over Bliss independence) � 1,2 > � 1 + � 2 – � 1 * � 2 Bliss CI. The toxicity of poisons applied jointly. Annals of Applied Biology 1939; 26:585–615.

Common Analysis Method Excess over Bliss Predicted � 1,2 = � 1 + � 2 – � 1 * � 2 , 0< � <1 Additivity : Observed � 1,2 – Predicted � 1,2 = 0 Synergism : Observed � 1,2 – Predicted � 1,2 > 0 Antagonism : Observed � 1,2 – Predicted � 1,2 < 0

Heat Map Using "Excess over Bliss” PI3K inhibitor BKM110 2 0.666667 0.222222 0.074074 0.024691 0.00823 0 µM 2 15 1 ‐ 2 0 ‐ 3 ‐ 11 0 0.666667 ‐ 13 0 0 8 0 ‐ 7 0 Gamitrinib 0.222222 ‐ 11 0 2 ‐ 7 ‐ 14 ‐ 27 0 0.074074 ‐ 9 4 3 25 15 8 0 0.024691 ‐ 12 14 21 13 5 2 0 0.00823 ‐ 12 9 7 25 8 13 0 0 0 0 0 0 0 0 0 Excess over Bliss scores = 0 indicates that the combination treatment is additive; Excess over Bliss scores < 0 indicates the combination is less than additive (antagonism); Excess over Bliss scores > 0 indicates activity greater than additive (synergism).

Other Analysis Methods Combination Index Chou TC and Talalay P. Generalized equations for the analysis of inhibitions of Michaelis ‐ Menten and higher ‐ order kinetic systems with two or more mutually exclusive and nonexclusive inhibitors. European Journal of Biochemistry . 1981; 115: 207 ‐ 216. Zhao, W., Sachsenmeier, K., Zhang, L., Sult, E., Hollingsworth. R. E., and Yang, H. (2014), “A new Bliss independence model to analyze drug combination data,” Journal of Biomolecular Screening, 19(5), 817–821. Whitehead, A., Su, T ‐ L., Thygesen, H., Sperrin, M., Harbron, C. (2013), “Investigation of the robustness of two models for assessing synergy in pre ‐ clinical drug combination studies,” Pharmaceutical Statistics, 12, 300 ‐ 308.

Two ‐ stage Nonlinear Dose ‐ Response Surface Model � � � � � � � � � � �1 � � � 0 �� � � � 0 � � � � 0 � � � � 0 ��� � �,� � �,� � �,� � �,� � � and � � are the doses of Drug A and B used in combination, reach the inhibition rate of � . � �,� and � �,� are the doses of single drugs A and B that reach the same inhibition rate, � . First stage: • Estimate the Hill Slope and IC50 for drugs A and B, respectively; • Bootstrapping the parameters estimated from stage I (random samples of Hill slope & IC50 drawn from bivariate-normal distribution for each drug with corresponding means & variance-covariance estimated from stage I); Second Stage: • Estimate Interaction Index Liu Q, Yin X, Languino LR, Altieri DC. Evaluation of drug combination effect using a Bliss independence dose-response surface model. Statistics in Biopharmaceutical Research . 2018. 10:2, 112-122, DOI: 10.1080/19466315.2018.1437071

Evaluation of drug combination effect - drugs in combination with Cisplatin Plate map for different drug concentrations combined with constant Cisplatin µ M 20 6.67 2.22 0.74 0.25 0.08 0.03 0.01 0.003 0.001 20 6.67 2.22 0.74 0.25 0.08 0.03 0.01 0.003 0.001 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 A DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO B DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO C SGC ‐ CBP30 PFI ‐ 3 DM SO DM SO Dox DM SO D SGC ‐ CBP30 PFI ‐ 3 DM SO DM SO Dox DM SO E I ‐ CBP112 C646# DM SO DM SO Dox DM SO F I ‐ CBP112 C646 DM SO DM SO Dox DM SO G DM SO DM SO JQ1/SGCBD01 LAQ824 Dox DM SO H JQ1/SGCBD01 LAQ824 DM SO DM SO Dox DM SO I bromosporine CI ‐ 994 DM SO Dox DM SO DM SO J DM SO Dox bromosporine CI ‐ 994# DM SO DM SO K GSK2801 SGC0946 DM SO Dox DM SO DM SO L GSK2801 SGC0946 DM SO Dox DM SO DM SO M PFI ‐ 1# LLY507 DM SO Dox DM SO DM SO N PFI ‐ 1 LLY507 DM SO Dox DM SO DM SO O DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO P DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO DM SO

Evaluation of Drug Combination Effect with Interaction Index

Validation of the Improved Efficacy of JQ1 and Cisplatin Combination in vivo Survival Tumor volume Yokoyama, …, Liu, …, Zhang. Inhibition of BET protein BRD4 activity synergizes with cisplatin in ovarian cancer by targeting ALDH activity through an ALDH1A1 super-enhancer and the associated enhancer RNA. Cancer Research . 2016

Collaborations During The Past 5 Years Two P01 Three U01 Several R01 and DOD grants > 80 coauthored scientific publications

Thank you!