Estimating the Causal Effects of T Cell Receptors (2410.14127v1)

Abstract: A central question in human immunology is how a patient's repertoire of T cells impacts disease. Here, we introduce a method to infer the causal effects of T cell receptor (TCR) sequences on patient outcomes using observational TCR repertoire sequencing data and clinical outcomes data. Our approach corrects for unobserved confounders, such as a patient's environment and life history, by using the patient's immature, pre-selection TCR repertoire. The pre-selection repertoire can be estimated from nonproductive TCR data, which is widely available. It is generated by a randomized mutational process, V(D)J recombination, which provides a natural experiment. We show formally how to use the pre-selection repertoire to draw causal inferences, and develop a scalable neural-network estimator for our identification formula. Our method produces an estimate of the effect of interventions that add a specific TCR sequence to patient repertoires. As a demonstration, we use it to analyze the effects of TCRs on COVID-19 severity, uncovering potentially therapeutic TCRs that are (1) observed in patients, (2) bind SARS-CoV-2 antigens in vitro and (3) have strong positive effects on clinical outcomes.

• The paper introduces a novel causal inference method using pre-selection TCR repertoires as instrumental variables to estimate TCR effects on patient outcomes.
• The paper employs CAIRE, a scalable neural network that efficiently processes high-dimensional, observational TCR sequencing data despite confounders.
• The paper demonstrates that specific TCR sequences are significantly linked to improved COVID-19 outcomes, suggesting promising targets for immunotherapy.

Estimating the Causal Effects of T Cell Receptors

In the presented paper, the authors introduce a method for inferring the causal effects of T cell receptor (TCR) sequences on patient outcomes. Their approach leverages observational TCR repertoire sequencing alongside clinical data, addressing the intrinsic challenge of unobserved confounders by employing the patient's immature, pre-selection TCR repertoire as an instrumental variable (IV). By employing nonproductive TCR data, which are generated by the V(D)J recombination process, the method draws causal inferences about the influence of TCRs on diseases such as COVID-19.

Methodological Contributions

1. Instrumental Variable Utilization: The cornerstone of this research is the use of pre-selection repertoires as an instrumental variable. These repertoires, unaffected by prior antigen exposure or confounding variables, provide an unbiased estimate of the initial distribution of TCR sequences. By using nonproductive TCR data as proxies for the pre-selection repertoire, the researchers mitigate issues related to confounding.
2. Neural Network Estimation: The authors develop CAIRE, a scalable neural network estimator for inferring causal effects. This deep learning approach allows for the handling of the high-dimensional and complex datasets typical of TCR repertoire sequencing.
3. Causal Identification: A formal identification strategy is laid out, demonstrating that under specified assumptions, the causal effects of adding specific TCR sequences to patient repertoires can be isolated and estimated.

Key Findings

• COVID-19 Analysis: As a practical application, the authors investigate the effects of TCRs on COVID-19 severity. They identify TCR sequences with significant positive impacts on clinical outcomes, which also show in vitro binding to SARS-CoV-2 antigens. This suggests potential therapeutic use.
• Immunological Implications: The outcomes support the notion that specific TCR interventions could inform the development of novel therapies, including TCR-T cell therapies and TCR vaccines.

Implications and Future Directions

This paper offers a significant advancement in understanding the causal relationship between TCR profiles and disease outcomes. The implications are substantial for designing personalized immunotherapies and enhancing the efficacy of existing treatments. Looking forward, such methodologies may be expanded to analyze B cell receptors, potentially impacting antibody-based treatment strategies. Additionally, refining causal estimation techniques to incorporate conditional treatment effects could stratify effects based on patient-specific characteristics, such as HLA types.

Conclusion

Through applying rigorous causal inference techniques to complex immunological data, this research paves the way for more precise and effective therapeutic interventions. With the evolving landscape of genomics and machine learning, such approaches will likely grow in relevance, potentially augmenting the predictive power and intervention efficacy in immunotherapy and beyond. By aligning computational methodologies with biological knowledge, the paper sets a precedent for future research in the intersection of machine learning and healthcare.