Predicting Immunotherapy Response: What Tumour Genetics Can Explain (and What It Cannot)?

Deficient mismatch repair (dMMR) is among the strongest predictive factors for ICI benefit in patients. It can lead to microsatellite instability (MSI), which increases the likelihood of accumulating mutations that may generate tumour-specific neoantigens [4]. These neoantigens, if effectively processed and presented via MHC-I molecules, can enhance T-cell recognition and anti-tumour immune responses [5]. This is supported by an in vivo study conducted by Peter et al. [6]. Using mouse models, they found that dMMR was associated with increased neoantigen burden, leading to high tumour mutational burden (TMB). Tumours with high TMB are often characterised by greater immune-cell infiltration, making tumour cells more susceptible to being recognised and attacked by the immune system. However, the study relied on relatively small sample sizes, and mouse models may not fully reflect the complex in vivo conditions in humans, limiting the representativeness of these findings.

Although dMMR is associated with high TMB, increased mutation load does not necessarily result in effective immune activation. Not all mutations generate immunogenic neoantigens, as only a small proportion of mutations can be efficiently processed and presented that trigger functional T-cell responses [7]. Moreover, chronic antigen exposure can induce persistent T-cell stimulation which may ultimately lead to T-cell exhaustion and limit therapeutic efficacy [8]. Tumours may also have developed various strategies to evade immune surveillance in the process of dMMR, leading to ultimate failure in ICIs, and this may drive more aggressive tumour progression and resistance to existing immunotherapy [9]. Taken together, these considerations indicate that while dMMR is a strong predictive biomarker, its predictive value remains probabilistic rather than deterministic, as other factors beyond tumour genetics ultimately shape treatment outcomes.

While tumour genetics can help predict effective ICIs responses, treatments with poor responses can also be predicted, one example is the deficiency in antigen presentation. Genetic or epigenetic alterations may disrupt key components of MHC-I antigen-processing and presentation machinery, leading to a reduction in the effectiveness of presentation of tumour-derived peptides on the cell surface. Such defects are associated with weaker immunotherapy responses under certain clinical contexts [10].

Yu et al. analysed secondary data from both patient samples and mouse models to support the idea that impaired MHC-I expression is significantly associated with reduced T-cell infiltration and poor ICI outcomes [11]. However, these findings are largely associative and rely partly on animal models, which limits firm conclusions about causality in human disease. Moreover, poor response cannot be fully explained by defective antigen presentation alone. The overall strength of the host immune system and the baseline composition of immune-cell populations also contribute to clinical outcomes. It is also challenging to measure tumour response reliably from a single tumour region, as antigen presentation can vary across different tumour regions and may evolve over time under immune pressure. Therefore, although antigen presentation deficiency represents an important mechanism of resistance, its predictive value remains context-dependent rather than a definitive rule.

Tumour genetics can provide valuable insights into immunotherapy response, but prediction sometimes fails, indicating that tumour genetics alone cannot fully account for clinical outcomes. This limitation mainly relates to tumour heterogeneity, the tumour microenvironment and host-level immune variability. Tumour genetics can differ substantially across tumour regions, meaning that a tumour biopsy from a single location may not accurately represent the overall mutational level. Thus, prediction based on one tumour sample may be a relatively unreliable predictor of the entire tumour behaviour in the body upon receiving immunotherapy.

The tumour microenvironment (TME) further shapes immunotherapy effectiveness and can limit the predictive value of tumour genetics. Cancer-associated fibroblasts (CAFs) are major components of the tumour stroma and can remodel the extracellular matrix (ECM) (e.g. collagen and matrix metalloproteinases, MMPs). The remodelling of the ECM may generate a dense stromal barrier that restricts infiltration of cytotoxic immune cells such as T cells and NK cells [12]. In addition to physical exclusion, the TME may also include several immunosuppressive cells including regulatory T cells, which may limit anti-tumour immunity [13]. In this situation, although tumour genetics may contain very sensitive genes which respond to immunotherapy, the physical barrier created by the ECM inside the tumour and potential immunosuppressive cells can prevent immune responses, leading to prediction failure.

Host-related factors also contribute to variability in treatment outcomes. Individuals have different immune status, due to baseline health, nutrient composition and prior treatments, and some might be immunocompromised. As a result, patients can respond differently to the same immunotherapy, even with similar tumour genetics and tumour types. Therefore, these factors illustrate why tumour genetics performs best when taking these various other factors into account rather than predicting with tumour genetics alone.

Prediction of immunotherapy response should integrate tumour genetics with additional biological and clinical determinants rather than relying on a single factor alone. This may involve combining dMMR status or impaired MHC-I antigen presentation with measures of immune-cell infiltration within the TME. In addition, immunotherapy may impose selection pressure on tumour cells, driving tumour evolution that may alter tumour genetics and treatment sensitivity over time. Because tumour genomics is dynamic and constantly changing rather than static, prediction based on a single sample is not enough for making reliable predictions. Instead, repeated assessments could be taken to track the progression of tumours and their response to immunotherapy. The combinations of different factors may provide a more robust predictive framework, which might reduce ineffective treatment, minimise unnecessary toxicity and high cost. As a result, this might improve overall clinical decision-making and lower the healthcare burden.