Can you explain the key components of Tempus Loop and how they’re integrated to form the platform?

Hagerty, Ph.D: We marry our AI-enabled RWD platform containing DNA, RNA, single-cell RNA-seq, spatial data, clinical data, knowledge graphs, and histopathology data with public datasets to identify multimodal cohorts and enable target identification through network mapping within a systems biology approach. We can then overlay these insights onto our extensive PDO models to validate our potential targets through various methods, including CRISPR knockdown screening and Perturb-seq.

What challenges in target identification and validation led to the development of Tempus Loop? How does it address these challenges?

Hagerty, Ph.D: Loop addresses two key challenges within target discovery. The first is access to a large collection of multimodal RWD, an area where Tempus is unparalleled. We can mine real-world patient data and align target discovery with true unmet patient needs. Data is key.

The second challenge is the fragmented nature of many drug discovery efforts, where teams often operate in silos and the steps from target identification to validation lack true continuity. Loop helps bridge this gap by creating a more integrated workflow.

How does the integration of multimodal RWD enhance the analytical power of Tempus Loop?

Hagerty, Ph.D: By combining our RWD with our modeling lab capabilities, Loop creates a unified approach that enables a ‘loop’, and an interactive hypothesis and validation engine, where drug discovery can enter at any point in this cycle.

What role do PDOs play in the target validation process within Tempus Loop?

Hagerty, Ph.D: Two key differentiators within our PDO models further the target validation process: 1) the PDOs are extensive and well-characterized, and 2) this characterization is aligned with how we integrate our multimodal clinical records. More importantly, we use the same xT and xR assays in both the organoids and within our RWD, which enhances the translation from bench to bedside.

How do AI and machine learning (ML) contribute to the target identification and validation process in Tempus Loop?

Hagerty, Ph.D: Our target discovery platform leverages advanced AI and ML techniques to enhance the identification and validation of gene targets associated with patient response. By training models with our extensive patient RWD, we can fine-tune transformer models to help pinpoint gene targets that correlate with therapeutic outcomes. Additionally, ML is instrumental in developing molecular classifiers that map PDO samples to patient molecular subtypes.

Furthermore, ML models are employed to identify gene targets associated with patient response and molecular subtypes. These models enable us to rank these targets for functional validation in PDO screens based on the important features identified by the models. This process ensures that the most promising gene targets are prioritized for further investigation.

How does Tempus Loop help ensure the accuracy and reliability of the data and insights it provides?

Hagerty, Ph.D: This is the beauty of Loop: we use our RWD to generate the potential targets, then we can test these within our PDOs, and then bring this back into the observed patient cohorts. Due to the breadth of our PDOs, we can map our subtypes to specific PDOs and identify control PDOs that do not have the same characteristics. Our CRISPR-based high-throughput screening allows for a quick assessment to determine potential candidates, which are then further assessed against control clusters and known oncogenic drivers.

What value does Tempus Loop bring to life science researchers and the broader scientific community?

Hagerty, Ph.D: First, speed. The integration of these approaches allows target identification and validation to occur more seamlessly, which can reduce the cycle time by half. Moreover, the ability to test further hypotheses against our multimodal patient data helps ensure that the targets identified and validated are more strongly linked to real-world clinical context.

Can you provide examples of how Tempus Loop has accelerated preclinical development in oncology research?

Hagerty, Ph.D: There are two recent examples that best illustrate the potential of Loop. The first used an unsupervised clustering approach to identify potential targets in NSCLC, which uncovered a novel patient cluster with a significant patient population and a high unmet clinical need. Leveraging our PDO biobank, we were able to quickly screen and validate these potential targets to narrow down a selection of targets that are unique and relevant for potential therapeutics.

The second example used a different entry point within our Loop process. In this case, we leveraged our pan-cancer PDO screening, testing standard-of-care, commercialized antibody-drug conjugates (ADCs), and their respective payloads across a range of PDOs. We tested enfortumab vedotin across these PDOs and created a unique gene signature correlating Nectin-4 expression to the response. We then brought this gene signature into our RWD and identified a cluster of patients who exhibited it. When compared to clinically treated patients without the gene signature, there was a significant difference in survival. In this example, we started with PDOs and carried the insights into the patient data, showing the translatability of our models.

What excites you the most about the future of Tempus Loop and its potential in advancing personalized cancer therapeutics?

Hagerty, Ph.D: As the next era of precision medicine continues to unfold, patient identification by a single protein expression may no longer be sufficient. The complexity of the disease and the impressive ability of cancer cells to adapt through the patient journey require a cohesive approach to target discovery and validation that addresses what is seen in the real-world patient populations.

If you could highlight one key takeaway for researchers considering using Tempus Loop, what would it be?

Hagerty, Ph.D: Relevance. We use patient data to drive the insights that are validated through the same characterization in our modeling lab, unifying target discovery and validation.