MSK opens postdoctoral post in computational immuno-oncology and immune toxicity research
Memorial Sloan Kettering Cancer Center is recruiting a postdoctoral fellow for a research program built around a problem that has become central to modern oncology: why immune-related adverse events emerge in some patients receiving cancer immunotherapy, why they remain confined to one organ in some cases, and why they spread into multisystem toxicity in others. The role sits inside the Reznik Laboratory within MSK’s Computational Oncology Service, where the work is framed at the intersection of cancer genomics, systemic physiology, and immuno-oncology.
The position is designed for a scientist who can move comfortably between high-dimensional data and clinical biology. Longitudinal clinical text, germline genetics, treatment exposures, laboratory measurements, and patient outcomes will all feed into the analytic program. In practical terms, that means the fellow will not be working with a single dataset in isolation. The assignment calls for integrating multiple data streams to reconstruct how immune toxicity begins, evolves, and diverges across patients exposed to immunotherapy.
A single adverse event can become a systems-level problem.
Linking toxicity phenotypes to the biology of immunotherapy
The scientific focus is more specific than a general interest in side effects. The laboratory wants to define, predict, and understand immune toxicity arising after cancer immunotherapy, with particular attention to the features that distinguish single-organ toxicity from multisystem involvement. That distinction matters biologically because immune checkpoint blockade and related treatments can alter self-tolerance in ways that are not uniform across patients. Some individuals may develop organ-restricted inflammation, while others may show a broader breakdown in immune regulation across tissues, suggesting different underlying susceptibility patterns.
The project framework leaves room for several analytic paths. One line of work centers on computational phenotyping using natural language processing and large language models. Another focuses on germline genetic analyses, including GWAS, HLA, and polygenic risk. A third uses multimodal modeling to search for predictors and possible biological mechanisms of toxicity. Each route addresses a different layer of the same clinical question: who develops immune toxicity, and what can the data reveal about why.
Because immune-related complications can appear late, fluctuate over time, or be documented unevenly in routine clinical care, longitudinal modeling is essential. Laboratory values, medication exposure, and narrative notes often contain the clues needed to distinguish a transient symptom from a clinically significant immune event. In a setting like this, computational methods are not decorative tools; they are the only practical way to recover signal from the volume and heterogeneity of the data.
Large-scale data, narrow clinical questions
The appointment is explicitly collaborative. The fellow will work with computational scientists, geneticists, oncologists, immunologists, and clinicians. That multidisciplinary structure reflects the nature of the problem itself. Toxicity from immunotherapy sits at the boundary of host genetics, treatment exposure, tumor biology, and immune activation, and no single analytic discipline can answer the full question on its own.
MSK describes the position as an opportunity to connect method development with clinically important questions about cancer immunotherapy. The scientific environment is built around rich longitudinal clinical and genomic data, which should allow the fellow to ask questions that are both technically sophisticated and directly relevant to patient care. For example, a model that identifies likely candidates for severe immune toxicity could influence monitoring intensity, prompt earlier recognition, or support future risk-stratification studies.
The role also signals where the field is moving. Immuno-oncology has expanded the therapeutic arsenal in cancer care, but the same immune activation that can produce durable antitumor responses can also injure normal tissue. The clinical challenge is not merely to catalog toxicity after it occurs. It is to anticipate it, classify it more accurately, and understand whether different toxicity patterns reflect different biology.
Methods expected from the next generation of computational oncologists
Candidates are expected to hold a PhD, completed or nearing completion, in a computational or quantitative field such as computational biology, bioinformatics, computer science, biomedical informatics, statistical genetics, biostatistics, statistics, engineering, or a related discipline. The desired background is broad but not undefined. Applicants should have expertise in at least one of several areas: textual analysis, natural language processing, or large language models; GWAS, statistical genetics, or germline genomics; or immune-related adverse events or immuno-oncology.
The laboratory is not asking for mastery of every method listed. Instead, it is looking for a scientist with strong quantitative reasoning, an interest in applying computational approaches to treatment toxicity, and the ability to develop rigorous, reproducible analyses from complex biomedical datasets. Independent project leadership is part of the expectation, but so is a willingness to contribute to collaborative work in a team that spans computational and clinical specialties.
Clear scientific writing also appears high on the list. In a setting where analytic output has to be interpreted by clinicians, geneticists, and methodologists, the ability to translate computational findings into usable biomedical language is not optional. It is central to the role.
What the appointment says about the future of toxicity research
The ad’s scientific emphasis places immune-related toxicity on the same footing as efficacy in immunotherapy research. For years, oncology has focused heavily on response rates, survival, and resistance biology. This position reflects a growing recognition that treatment harm has its own mechanistic architecture and deserves the same level of analytic rigor.
The work is likely to be especially relevant for studies that try to separate shared inflammatory pathways from organ-specific signatures. HLA associations, polygenic risk, and treatment exposure patterns may help explain why some patients remain unaffected while others develop clinically meaningful immune injury. Natural language processing can further improve case identification by mining the clinical record for patterns that structured fields often miss.
MSK sets the pay range at 72,000 to 75,000 dollars and identifies the site as New York City. The closing date is listed as 27 December 2026. The application instructions request a CV, letter of interest, and references, and the post is framed as part of a broader institutional commitment to training early-career scientists within a major cancer center.
For investigators interested in immunotherapy toxicity, the unresolved issue is not whether these events matter clinically. They do. The harder problem is to convert longitudinal clinical data and germline variation into a mechanism-aware model of risk that can hold up across patients, treatments, and organ systems.