Abstract:Lymphoma and leukemia have been the top two hematopoietic cancers in US, leading to more than 40,000 deaths each year. Although the 5-year relative survival rates have been improved up to 60%, there is still a lot of patients showed not response to standard therapy. Patient-Derived Xenograft model (PDX model), as the classical mouse xenograft tumor model that best represents the genetic information characteristics of human tumors, can be used to predict the therapeutic effect, and develop individualized treatment for patients. Its greatest advantage is that the model retains the microenvironment of the original tumor cells, inherits all the molecular biological properties of the primary tumor, and preserves the heterogeneity of the tumor. However, due to the factors that the loss of tumor heterogeneity in hematopoietic cancers after ex vivo clonal expansion and selection, and this in vitro assay based system could not fully reflect the host response to the drugs, including a range of novel immunotherapeutic agents such as PD-1 monoclonal antibody, CD38 monoclonal antibody, etc. With the increasing clinical use of oncology immunological agents, recently we are developing a new assay, named IO-FIVE (Immuno-Oncology drugs Fast In Vivo Efficacy test) for research and therapy, by retaining an appropriate proportion of tumor-infiltrating immune cells (and stromal cells) mixed with tumor cells when digesting clinical tumor samples or patient-derived xenografts-preserved tumor samples into cell suspensions, and then injecting them into the specialized IO-FIVE device, a modified microencapsulation and hollow fiber culture system (OncoVee® MiniPDX), which is implanted subcutaneously into mice and administered systematically for immuno-drug susceptibility testing to screen for immuno-drugs or combinations suitable for individual treatment. The IO-FIVE testing cycle takes only 14 days and is performed using Celltiter Glo before and after the injection of cell suspensions into the specialized device. Celltiter Glo, flow cytometry and Omics (RNA-seq, DNA-seq) are used to measure the viability of total cell subpopulations, the relative ratio of tumor cells to immune cells and the alteration of cellular transcription levels in the device, in order to further explore the sensitivity of tumor tissues to the immunotherapeutic drugs and the potential molecular mechanism for the responders and non-responders. Immune regulatory CD38 antibody Daratumumab has been tested in more than 40 acute myeloid leukemia (AML) cases, with its overall in vivo efficacy around 25%, which is consistent with previous research. It is interesting to uncover novel pathways for drug discovery and clinic treatment since CD38 is well expressed in most AML patients, but at least half of those CD38 + AML patients are not sensitive or resistant to Daratumumab monotherapy. Expression of CD38 on AML is not enough to distinguish the responder vs non-responder of Daratumumab. IO-FIVE could potentially be a companion diagnosis on AML for the patient stratification of Daratumumab. PD1 antibody Sintilimab has been tested in all the patients (12 lymphoma, 52 leukemia), and the only two laboratory PD1-Ab non-responsive T cell lymphoma patients have been showed to be clinically non-responders as well. Immunotherapy and chemical or targeted therapy combination could enhance tumor killing ability and disease control. We will use the IO-FIVE data to further compare it with its source of real-world clinical patients to obtain more evidence-based evidence that can predict individualized treatment with immunotherapy.

1O Uncovering New Immunotherapy Targets with Machine Learning and Ex Vivo Validation

Abstract 5887: Identifying new immunotherapy targets using machine learning and ex vivo validation

Informing immunotherapy with multi-omics driven machine learning

Quantitative multiplexed imaging technologies for single-cell analysis to assess predictive markers for immunotherapy in thoracic immuno-oncology: promises and challenges

Facts and hopes in using omics to advance combined immunotherapy strategies

Using Machine Learning Algorithms to Predict Immunotherapy Response in Patients with Advanced Melanoma

Ex vivo modeling of precision immuno-oncology responses in lung cancer

Multi-omics and artificial intelligence predict clinical outcomes of immunotherapy in non-small cell lung cancer patients

Cancer immunotherapy efficacy and machine learning

A Novel Patient-Derived Xenograft Model IO-FIVE for Drug Discovery and Precision Medicine in Lymphoma and Leukemia Patients

Multimodal machine learning model prediction of “individual” response to immunotherapy in 1L stage IV NSCLC.

Uncovering Predictive Gene and Cellular Signatures for Checkpoint Immunotherapy Response through Machine Learning Analysis of Immune Single-Cell RNA-seq Data

Machine Learning for Prediction of Immunotherapy Efficacy in Non-Small Cell Lung Cancer from Simple Clinical and Biological Data

Microfluidics guided by deep learning for cancer immunotherapy screening

Network-based machine learning approach to predict immunotherapy response in cancer patients

Biomarkers and computational models for predicting efficacy to tumor ICI immunotherapy

Historical perspective and future directions: computational science in immuno-oncology

Abstract 117: Improved PBMC humanized mouse models for immuno-oncology drugs evaluation

Exploration of the Underlying Biological Differences and Targets in Ovarian Cancer Patients with Diverse Immunotherapy Response

Multi‐omic and spatial dissection of immunotherapy response groups in non‐small cell lung cancer

Immunotherapy Efficacy Prediction in Cancer: An Artificial Intelligence Approach with Unannotated H&E Whole-Slide Images