AI and Immunotherapy: How Artificial Intelligence Is Transforming Precision Cancer Treatment
Medically Reviewed by: OneDayMD Editorial Team | Written by: Dr Frank Yap, MD | Last Updated: August 8, 2026
Artificial intelligence (AI) and cancer immunotherapy are converging to create a new generation of precision oncology. AI can analyze enormous amounts of cancer data—including pathology images, genomic profiles, blood tests, clinical records, radiology and tumor-microenvironment measurements—to identify patterns that may help predict which patients are most likely to benefit from immunotherapy.
This is particularly important because immunotherapy does not work equally well for every patient. Even established biomarkers such as PD-L1 expression and tumor mutational burden (TMB) have important limitations. Some patients with apparently unfavorable biomarkers respond dramatically, while some patients with favorable biomarkers fail to benefit.
.png)
AI may help address this problem by moving oncology beyond single biomarkers toward multimodal prediction—integrating multiple biological and clinical signals into a more comprehensive assessment of tumor-immune interactions.
What Is AI-Powered Immunotherapy?
AI-powered immunotherapy refers to the use of machine learning, deep learning and other computational methods to improve the development, selection, monitoring or personalization of cancer immunotherapies.
AI can potentially analyze information from several levels of cancer biology simultaneously:
- Genomics: mutations, neoantigens and tumor mutational burden
- Transcriptomics: gene-expression patterns and immune signatures
- Proteomics: proteins and signaling pathways
- Digital pathology: tumor architecture and immune-cell distribution
- Radiology: CT, MRI and PET imaging features
- Blood biomarkers: immune cells, inflammatory markers and circulating tumor signals
- Clinical data: age, cancer type, previous treatments and laboratory measurements
- Tumor microenvironment: immune-cell infiltration, stromal characteristics and spatial relationships
The objective is to transform these high-dimensional datasets into clinically useful predictions.
Why AI Is Needed in Cancer Immunotherapy
Immunotherapy is fundamentally different from many conventional cancer treatments because its effectiveness depends not only on the cancer cell itself but also on the interaction between the tumor and the patient's immune system.
The tumor microenvironment can contain:
- T cells
- Natural killer cells
- Macrophages
- Dendritic cells
- Regulatory T cells
- Myeloid-derived suppressor cells
- Fibroblasts
- Cytokines and chemokines
- Blood vessels
- Extracellular matrix components
These components interact dynamically. A single biomarker may therefore provide only a partial picture.
AI is attractive because machine-learning algorithms can identify complex relationships across thousands or millions of variables that would be difficult to evaluate manually.
AI and PD-L1: Going Beyond a Single Biomarker
PD-L1 is one of the best-known biomarkers used in cancer immunotherapy. Higher PD-L1 expression can be associated with a greater probability of response to certain immune checkpoint inhibitors, depending on cancer type and treatment setting.
However, PD-L1 is not a perfect predictor.
A patient with low PD-L1 expression may still respond to immunotherapy, while another patient with high PD-L1 expression may not.
This is one reason researchers are developing AI models that incorporate multiple variables rather than relying on PD-L1 alone.
The National Cancer Institute has highlighted this challenge and the development of computational approaches that combine clinical, molecular and pathological information to improve prediction of immunotherapy outcomes.
AI and Tumor Mutational Burden (TMB)
Tumor mutational burden (TMB) measures the number of mutations found within a tumor's DNA. Because mutations can generate abnormal proteins, a high mutation burden may increase the probability that the immune system can recognize cancer cells.
TMB can therefore provide useful information when considering immune checkpoint inhibitors.
But, like PD-L1, TMB is not a universal predictor of immunotherapy response.
Interestingly, an NCI-developed machine-learning model called LORIS was able to identify patients who might benefit from immune checkpoint inhibitors even among patients with relatively low TMB. The model incorporated TMB together with routinely collected clinical information.
LORIS: AI Prediction From Routine Clinical Data
One of the important developments in AI immuno-oncology is that sophisticated models do not necessarily require enormous amounts of expensive genomic data.
The NCI-supported Logistic Regression-Based Immunotherapy-Response Score (LORIS) uses routinely available clinical information together with TMB.
The model incorporated factors including:
- Patient age
- Cancer type
- Previous systemic therapy
- Blood albumin level
- Neutrophil-to-lymphocyte ratio
- Tumor mutational burden
The original analysis included data from 2,881 patients receiving immune checkpoint inhibitors across 18 solid tumor types. The researchers reported that the model could predict the likelihood of response and survival outcomes, but emphasized that larger prospective studies are needed before such approaches can become routine clinical decision tools.
SCORPIO: Using Routine Blood Tests to Predict Immunotherapy Benefit
Another important development is SCORPIO, an AI model designed to predict outcomes from immune checkpoint inhibitors using routine clinical and laboratory information.
According to the NCI, SCORPIO was designed to use information that is already commonly collected in clinical practice, potentially making AI-assisted immunotherapy prediction more accessible than approaches requiring sophisticated molecular testing.
This illustrates an important direction for AI oncology: clinically useful AI does not necessarily have to depend on exotic or expensive datasets.
AI Can Analyze the Tumor Microenvironment
One of the most promising applications of AI is analysis of the tumor microenvironment (TME).
The TME contains cancer cells, immune cells, blood vessels, connective tissue and signaling molecules. The composition and spatial organization of these components can influence whether an immune response develops inside a tumor.
AI can analyze digital pathology slides to identify patterns involving the distribution and interaction of different cell populations.
For example, the NCI highlighted HistoTME, an AI model developed to analyze routine pathology images and infer characteristics of the tumor microenvironment. In research involving non-small-cell lung cancer, the model was investigated as a way of predicting tumor characteristics relevant to immunotherapy response.
AI and Digital Pathology
Traditional pathology depends heavily on expert interpretation of tissue morphology. Digital pathology converts these slides into high-resolution images that can be analyzed computationally.
AI can potentially identify:
- Immune-cell density
- T-cell infiltration
- Spatial relationships between immune cells and cancer cells
- Tumor architecture
- Necrosis
- Stromal patterns
- Features associated with treatment resistance
This creates a potential bridge between conventional pathology and computational oncology.
AI and Multi-Omics Immunotherapy Prediction
The next generation of AI models is likely to integrate multiple biological layers simultaneously.
A hypothetical multimodal immunotherapy model could combine:
- PD-L1 expression
- TMB
- MSI status
- Genomic mutations
- Gene-expression signatures
- Immune-cell infiltration
- Digital pathology
- Radiomics
- Blood biomarkers
- Clinical characteristics
- Previous treatment history
Instead of asking "Does this tumor have PD-L1?", the system could eventually address a more clinically meaningful question:
Recent reviews describe this movement toward multimodal AI models integrating genomics, transcriptomics, radiomics, digital pathology, circulating biomarkers and clinical information.
AI for Immunotherapy Resistance
One of the biggest challenges in oncology is immunotherapy resistance.
Some tumors never respond to checkpoint inhibitors. Others initially respond but eventually develop resistance.
Potential mechanisms include:
- Loss or alteration of tumor antigens
- Defects in antigen presentation
- Changes in interferon signaling
- Immunosuppressive tumor microenvironments
- Exhausted T cells
- Changes in immune-cell composition
- Activation of alternative immune checkpoints
- Tumor evolution under treatment pressure
AI may help researchers discover combinations of biomarkers associated with these resistance mechanisms and identify patients who may require alternative or combination strategies.
AI and Immune Checkpoint Inhibitors
Immune checkpoint inhibitors—including therapies targeting PD-1, PD-L1 and CTLA-4—have transformed treatment for multiple cancers.
AI research is exploring several questions:
- Who is most likely to respond?
- Who is unlikely to benefit?
- Which patients may experience durable responses?
- Who is at increased risk of immune-related adverse events?
- Which combinations may overcome resistance?
- When should treatment be changed?
The potential value is substantial because immunotherapy can produce long-lasting responses in some patients, while exposing others to treatment costs and potentially serious toxicities without meaningful benefit.
AI and Immune-Related Adverse Events
Immunotherapy can cause immune-related adverse events (irAEs) because stimulating the immune system can sometimes cause inflammation in healthy organs.
Potential complications can involve:
- Skin
- Colon
- Liver
- Lungs
- Endocrine organs
- Kidneys
- Heart
- Nervous system
AI may eventually help identify patients at higher risk of severe toxicity by analyzing combinations of clinical, laboratory and biological variables.
Current research reviews identify toxicity prediction and monitoring as important applications of AI in cancer immunotherapy, although much of this work remains investigational.
AI for Personalized Cancer Vaccines
Another major application is the development of personalized cancer vaccines.
Some tumors contain mutations that generate abnormal proteins known as neoantigens. If these neoantigens can be recognized by the immune system, they may provide targets for personalized immunotherapy.
AI can potentially assist researchers by analyzing tumor sequencing data and predicting which mutations are most likely to produce immunogenic neoantigens.
This could accelerate the process of identifying candidate targets for individualized vaccines or T-cell-based therapies.
AI and CAR-T Cell Therapy
AI is also being investigated in cellular immunotherapy.
Potential applications include:
- Identifying tumor-specific antigens
- Predicting T-cell activity
- Designing T-cell receptors
- Optimizing CAR structures
- Predicting treatment response
- Identifying mechanisms of resistance
- Monitoring cellular therapy
NCI researchers have already used machine learning to study patterns associated with T-cell activation and behavior, demonstrating how computational approaches can support the development of next-generation immunotherapies.
AI and "Hot" vs "Cold" Tumors
One of the central concepts in immuno-oncology is the distinction between immune-hot and immune-cold tumors.
Immune-hot tumors generally contain greater immune-cell infiltration and may be more susceptible to immune-mediated attack. Immune-cold tumors may have limited immune infiltration or other characteristics that prevent effective immune activation.
AI may help classify tumors based on complex combinations of pathological, molecular and spatial characteristics rather than relying on a single marker.
This could eventually support strategies designed to convert immune-cold tumors into more immunologically active tumors.
AI + Radiomics + Immunotherapy
Radiomics involves extracting quantitative features from medical images such as CT, MRI or PET scans.
AI can analyze subtle imaging characteristics that may not be obvious to the human eye.
Potential applications include:
- Predicting immunotherapy response
- Monitoring tumor evolution
- Detecting early treatment response
- Identifying heterogeneous tumor regions
- Predicting progression
Combining radiomics with pathology, genomics and blood biomarkers may ultimately produce a more complete representation of tumor biology.
AI and Immunotherapy: The Emerging Multimodal Model
The most promising direction is not necessarily one "AI biomarker."
Instead, oncology may move toward multimodal AI.
The AI Immunotherapy Data Stack
- Clinical: age, performance status, previous treatments
- Blood: inflammatory and immune markers
- Genomics: mutations, TMB and MSI
- Transcriptomics: gene-expression signatures
- Pathology: tumor and immune-cell architecture
- Radiology: CT, MRI and PET features
- Tumor microenvironment: immune and stromal characteristics
- Longitudinal data: changes during treatment
The model can then generate a probability or risk estimate rather than relying on a single yes/no biomarker.
Could AI Eventually Choose the Best Immunotherapy?
Potentially—but this remains an emerging research area.
A future clinical decision-support system could theoretically compare multiple treatment strategies and estimate:
- Probability of response
- Probability of durable response
- Risk of severe toxicity
- Probability of resistance
- Expected progression-free survival
- Potential benefit of combination therapy
However, an AI prediction should not automatically be interpreted as a treatment recommendation.
Clinical decisions still require consideration of the patient's overall condition, cancer stage, pathology, biomarkers, previous treatments, comorbidities, drug interactions, treatment guidelines and patient preferences.
AI Is Not Yet a Replacement for Clinical Trials
This distinction is critical.
A model can achieve impressive accuracy in retrospective datasets and still fail when deployed in a different hospital, population or cancer type.
Common problems include:
- Overfitting
- Selection bias
- Incomplete clinical data
- Dataset shift
- Differences between hospitals
- Differences in pathology scanners
- Differences in treatment protocols
- Limited representation of minority populations
- Lack of prospective validation
Consequently, AI models that predict immunotherapy response need rigorous external validation and, where appropriate, prospective clinical evaluation.
AI in Cancer: Prediction vs Proof
One of the most important principles for patients is to distinguish between prediction and clinical proof.
An AI model may identify a patient as having a high probability of responding to immunotherapy. That does not prove that the treatment will work.
Similarly, a low AI score does not necessarily mean that immunotherapy will fail.
AI should therefore be viewed as another layer of evidence—not as an infallible diagnostic oracle.
What Is the Evidence for AI and Immunotherapy in 2026?
The evidence base has moved beyond purely theoretical applications.
By 2026, researchers have demonstrated AI approaches using routine blood data, clinical information, pathology images and molecular data to predict immunotherapy outcomes. The NCI has highlighted LORIS, SCORPIO and HistoTME as examples of this rapidly developing field.
A July 2026 Nature Medicine study further reported a generalizable AI approach for predicting immunotherapy outcomes across cancers and treatments, illustrating the field's movement toward models designed to generalize across multiple cancer types and therapeutic settings.
At the same time, the field remains developmental. Reviews published in 2025–2026 emphasize the potential of AI for biomarker discovery, response prediction, treatment personalization and toxicity monitoring while noting the need for further validation and clinical translation.
The Future: An AI Immunotherapy Copilot
The most realistic near-term vision is not an autonomous AI oncologist.
It is an AI immunotherapy copilot that helps oncologists integrate information that is too complex to evaluate efficiently by conventional methods alone.
Such a system could potentially:
- Analyze pathology slides
- Interpret genomic profiles
- Integrate PD-L1 and TMB
- Evaluate immune-cell infiltration
- Analyze longitudinal blood tests
- Assess radiological changes
- Estimate treatment response
- Identify potential resistance patterns
- Flag possible immune-related toxicity
- Match patients to relevant clinical trials
The FDA has established an Oncology Artificial Intelligence Program to advance understanding and regulatory science around AI applications in oncology drug development, reflecting the increasing importance of these technologies in cancer research and development.
AI + Immunotherapy: What Patients Should Know
If you or someone you know is considering immunotherapy, AI can potentially provide additional information, but conventional cancer evaluation remains essential.
Important factors may include:
- Cancer type and stage
- Histological diagnosis
- PD-L1 status where clinically relevant
- TMB
- MSI/MMR status
- Relevant genomic alterations
- Performance status
- Previous treatments
- Overall health
- Potential contraindications
- Clinical trial availability
Biomarker testing can help physicians determine whether specific treatments may be appropriate. The NCI notes that biomarker testing can examine genes, proteins and other characteristics of a patient's cancer and may help guide treatment selection.
Frequently Asked Questions
Can AI predict whether immunotherapy will work?
AI can estimate the probability of response, but it cannot guarantee that immunotherapy will work. Several research models have demonstrated promising predictive performance, but prospective validation and clinical implementation remain important.
Can AI replace PD-L1 testing?
No. AI does not currently replace established clinical biomarkers. Instead, AI may combine PD-L1 with other clinical, molecular, pathological and imaging information to improve prediction.
Can AI predict immunotherapy side effects?
Researchers are investigating AI models that predict immune-related adverse events and treatment toxicity. These applications remain an active research area.
Can AI identify the best immunotherapy?
AI may eventually help compare treatment options, but it should currently be regarded as decision-support technology rather than an autonomous treatment-selection system.
Can AI discover new cancer immunotherapy targets?
Yes. AI and machine learning are being used to analyze biological datasets, identify potential targets, study immune-cell behavior and support drug discovery.
Is AI-based immunotherapy available to patients today?
Some AI technologies are already being used in oncology research and selected clinical applications, but many AI models specifically designed to predict immunotherapy response remain investigational rather than established standards of care.
Bottom Line
AI and immunotherapy represent one of the most important emerging intersections in precision oncology.
The central opportunity is to move from simplistic biomarker-based treatment selection toward a more comprehensive understanding of each patient's tumor, immune system and treatment history.
AI can potentially integrate PD-L1, TMB, genomics, pathology, radiology, blood biomarkers, tumor microenvironment data and clinical history into individualized predictions of immunotherapy response and toxicity.
The science is advancing rapidly. Research involving LORIS, SCORPIO, HistoTME and newer multimodal models demonstrates that AI can extract clinically relevant signals from data that are already generated during cancer care.
But the most responsible conclusion in 2026 is that AI is an emerging tool for precision immuno-oncology—not a replacement for clinical evidence, validated biomarkers, clinical trials or oncologists.
Sources & Further Reading
- National Cancer Institute — AI and Cancer
- National Cancer Institute — LORIS and AI prediction of immunotherapy response
- National Cancer Institute — SCORPIO and blood-based prediction of immunotherapy outcomes
- National Cancer Institute — HistoTME and tumor-microenvironment analysis
- Nature Medicine — Generalizable AI predicts immunotherapy outcomes across cancers and treatments
- Nature Communications — Decoding immunotherapy response through computational modeling
- Frontiers in Immunology — Applications of artificial intelligence in cancer immunotherapy
Medical disclaimer: This article is for educational and informational purposes only and does not constitute medical advice, diagnosis or treatment. Cancer treatment decisions should be made with qualified oncology professionals.
.png)
.png)
Comments
Post a Comment