Early identification of immunotherapy response based on computational models

Early identification of immunotherapy response based on computational models

Immune checkpoint inhibition is a new pillar in cancer treatment that has earned the 2018 Nobel Price in "Physiology or Medicine". Tumor cells "shut off" lymphocyte attacks by activating "immune checkpoint" molecules on these immune cells. The most important of these checkpoints is the interaction of the surface molecule PD1 (programmed death 1) with its ligand PDL1. Antibodies that inhibit this interactions are called "Anti-PD1/PDL1" and can reinvigorate the immune response in certain cancer types. Another type of immune checkpoints is the cell surface molecule "CTLA4" and inhibitors of this checkpoint molecule ("Anti-CTLA4-antibodies") can likewise induce an immune response against the cancer cells. In summary, this type of treatment can induce durable responses in a subset of cancer patients.However, in the clinic, it is a common problem to identify these responders early during treatment. The temporal dynamics of response to anti-PD1/PDL1 treatment are different than for classical cytotoxic or targeted agents: Some patients respond immediately, some patients show a pseudoprogression followed by a response and some patients stay stable for a prolonged time and respond after months of treatment.In our interdisciplinary group, we have previously developed computational models that can predict immunotherapy response in individual patients. These models can be calibrated to individual patients and can predict optimal treatment schedules and combinations (Kather et al., Cancer Research 2017 and Kather et al., Cancer Research 2018; PMID 29967263 and 28923860). These previously published models relied on spatial data of cells in the tumor microenvironment (TME) obtained via image analysis of histological slides. These models are spatial simulations of tumors and require extensive histological data. Because this data is not always available, we have developed another computational modeling framework that requires less data. This new model is a non-spatial differential-equation-based model which can be used for response prediction and mechanistic optimization of cancer treatments, as explained below.In short, these non-spatial computational models aim to predict response and allow scheduling of optimal treatment plans based on non-spatial data. In our preliminary experiments we have used these computational models in small cohorts of patients treated with immunotherapy in different tumor types. Our preliminary data show that these models can1) identify non-response and response early during treatment2) identify non-responders to anti-PD1/PDL1 treatment who could be rescued by adding anti-CTLA4 antibodies3) predict the lymphocyte number in the tumor microenvironment non-invasivelyThis non-spatial model requires de-identified patient level data (tumor volume measurements over time, clinical outcome and, optionally, lymphocyte density in tumor tissue) as explained in section 3.1.We are planning to validate our model retrospectively in patient cohorts who were treated with anti-PD1/PDL1 treatment and we kindly request access to this anonymized, readily available data from the selected previous clinical trials. The ultimate aim of our proposal is to build the infrastructure for the next generation of clinical trials. The ultimate aim of our proposal is to build the infrastructure for the next generation of clinical trials. This will directly benefit patients because the current landscape of clinical trials has some key limitations:1. Clinical trials inherently operate on a group level and usually do not incorporate personalized predictions2. Clinical trials generally test only one or two treatment strategies. We can expect that shortly, dozens of immunotherapies will be available and testing of all possible combinations is not feasible.Our proposal will benefit patients because it will allow us to develop the technological basis for a mechanistic model that addresses both of these points.

[{ "PostingID": 14579, "Title": "ROCHE-GO28915", "Description": "A Study of Atezolizumab Compared With Docetaxel in Participants With Locally Advanced or Metastatic Non-Small Cell Lung Cancer Who Have Failed Platinum-Containing Therapy" },{ "PostingID": 14580, "Title": "ROCHE-GO28625", "Description": "A Study of Atezolizumab in Participants With Programmed Death-Ligand 1 (PD-L1) Positive Locally Advanced or Metastatic Non-Small Cell Lung Cancer (NSCLC) [FIR]" },{ "PostingID": 14581, "Title": "ROCHE-GO28754", "Description": "A Study of Atezolizumab in Participants With Programmed Death - Ligand 1 (PD-L1) Positive Locally Advanced or Metastatic Non-Small Cell Lung Cancer" },{ "PostingID": 14582, "Title": "ROCHE-GO28753", "Description": "A Randomized Phase 2 Study of Atezolizumab (an Engineered Anti-PDL1 Antibody) Compared With Docetaxel in Participants With Locally Advanced or Metastatic Non-Small Cell Lung Cancer Who Have Failed Platinum Therapy - \"POPLAR\"" },{ "PostingID": 16140, "Title": "ROCHE-GO29293 (cohort 1)", "Description": "A Study of Atezolizumab in Participants With Locally Advanced or Metastatic Urothelial Bladder Cancer (Cohort 1)" },{ "PostingID": 16142, "Title": "ROCHE-GO29293 (cohort 2)", "Description": "A Study of Atezolizumab in Participants With Locally Advanced or Metastatic Urothelial Bladder Cancer (Cohort 2)" }]

Classical Mathematical Models for Prediction of Response to Chemotherapy and Immunotherapy
Narmin Ghaffari Laleh, Chiara Maria Lavinia Loeffler, Julia Grajek, Kateřina Staňková, Alexander T. Pearson, Hannah Sophie Muti, Christian Trautwein, Heiko Enderling, Jan Poleszczuk, Jakob Nikolas Kather
bioRxiv 2021.10.23.465549; doi: https://doi.org/10.1101/2021.10.23.465549