Predicting Likelihood of In Vivo Chemotherapy Response in Canine Lymphoma Using Ex Vivo Drug Sensitivity and Immunophenotyping Data in a Machine Learning Model
2020 VCS Virtual Conference
Zach Bohannan1; Raghavendra Pudupakam1; Jamin Koo1,2,3; Harrison Horwitz1; Josephine Tsang1; Amanda Polley1; James Enyang Han1; Elmer Fernandez1; Stanley Park1; Deanna Swartzfager1; Nicholas Seah Xi Qi1; Chantal Tu4; Wendi Velando Rankin4; Douglas H. Thamm5; Hye-Ryeon Lee1,2; Sungwon Lim1,2
1ImpriMed, Inc., USA; 2ImpriMed Korea, Inc., Seoul, Republic of Korea; 3Department of Chemical Engineering, Hongik University, Republic of Korea; 4SAGE Veterinary Centers, USA; 5Flint Animal Cancer Center, USA

Introduction

For a life-threatening disease like cancer, optimal treatment may need to be personalized for each patient. We developed a precision medicine platform that evaluates the probability of chemotherapeutic efficacy for canine lymphoma by combining ex vivo chemosensitivity and immunophenotyping assays with computational modeling. This study evaluated the correlation between our ex vivo data-driven prediction and reported responses in a series of canine lymphoma patients.

Methods

We collected live cancer cells from fresh fine needle aspirates taken from affected lymph nodes and collected post-treatment clinical responses in 261 canine lymphoma patients who were scheduled to be treated with at least 1 of 5 common chemotherapy agents (doxorubicin, vincristine, cyclophosphamide, lomustine, and rabacfosadine). Samples were subject to flow cytometry analysis for immunophenotyping and ex vivo chemosensitivity testing. For each drug, 70% of the treated patients were randomly selected to train a random forest model to predict the probability of positive Veterinary Cooperative Oncology Group (VCOG) clinical response based on input variables including antigen expression profiles and treatment sensitivity readouts for each patient’s cancer cells. The remaining 30% of patients were used to test model performance.

Results

Most models showed good performance by ROC-AUC and Brier score. Predictive efficacy scores significantly distinguished (p<0.0001) positive responses from negative responses, both when comparing B- vs. T-cell patients and naïve vs. relapse patients. In addition, the patient group with predictive efficacy scores >50% showed a statistically significant reduction (log-rank p<0.05) in time to complete response when compared to the group with scores <50%.

Conclusions

The computational model developed in this study enabled conversion of ex vivo cell-based assay results into probability of in vivo therapeutic efficacy, which may help improve treatment outcomes in individual canine lymphoma patients by providing predictive estimates of positive treatment response.

 

Speaker Information
(click the speaker's name to view other papers and abstracts submitted by this speaker)

Lim Sungwon
ImpriMed, Inc.


MAIN : International Abstracts : Predicting Likelihood of Chemotherapy Response
Powered By VIN
SAID=27