The rise of cellular immunotherapy for cancer treatment has led to the utilization of immune oncology cocultures to simulate T cell interactions with cancer cells for assessing their antitumor response. Previously, we developed BEHAV3D, a three-dimensional live imaging platform of patient-derived tumor organoid (PDO) and engineered T cell cocultures, that analyzes T cells' dynamics to gain crucial insights into their behavior during tumor targeting. However, live imaging alone cannot determine the molecular drivers behind these behaviors. Conversely, single-cell RNA sequencing (scRNA-seq) allows researchers to analyze the transcriptional profiles of individual cells but lacks spatio-temporal resolution. Here we present an extension to the BEHAV3D protocol, called Behavior-Guided Transcriptomics (BGT), for integration of T cell live imaging data with single-cell transcriptomics, enabling analysis of gene programs linked to dynamic T cell behaviors. BGT uses live imaging data processed by BEHAV3D to guide the experimental setup for cell separation based on their PDO engagement levels subsequently followed by fluorescence-activated cell sorting and scRNA-seq. It then integrates in silico simulations of these experiments to computationally infer T cell behavior on scRNA-seq data, uncovering new biomarkers for both highly functional and ineffective T cells, that could be exploited to enhance therapeutic efficacy. The protocol, designed for users with fundamental cell culture, imaging and programming skills, is readily adaptable to diverse coculture settings and takes one month to perform.