Physical models of immunotherapy

Valentinuzzi, Damijan (author), Jeraj, Robert (mentor)

Immunotherapy with immune-checkpoint inhibitors has changed the paradigm of cancer treatment. For the first time in the history of oncology, it is possible to realistically expect complete remissions of metastatic disease in certain types of cancers such as melanoma and non-small-cell lung cancer (NSCLC), which were previously considered almost incurable. However, patients with such favourable responses are the minority, and the reasons remain mostly unknown. The overall goal of this thesis was to identify tumour characteristics that drive patient response to anti-programmed-death-1 (anti-PD-1) immunotherapy, using computational modelling. To examine this objective, we exploited two opposite modelling approaches: a mechanistic bottom-up approach and a data-driven top-down approach. Our bottom-up modelling was focused on pre-clinical anti-PD-1 experiments to ensure sufficient experimental data, needed for model tuning and validation. Two similar models were developed, which were tuned and validated on murine data from literature and data obtained by on-site experiments, respectively. Both models identified tumour major histocompatibility complex (MHC) class I expression as a biomarker of resistance to anti-PD-1 therapy. Moreover, they suggested that complete responses to anti-PD-1 can occur only in the case of homogeneous (MHC class I positive) tumours. On the other hand, a data-driven top-down model was developed to examine a clinical study containing sparse data, where metastatic NSCLC patients were treated with pembrolizumab and scanned with [18F]FDG PET/CT. This model was based on radiomics analyses of PET scans of primary tumours (iRADIOMICS) and was found predictive of patient overall survival. Interestingly, a detailed inspection of model components suggested that tumour homogeneity is associated with favourable response to anti-PD-1 immunotherapy, similar to the bottom-up approach findings. If findings arising from this thesis are confirmed in independent studies, they will have a significant impact on the clinical practice, allowing for better management of cancer patients treated with anti-PD-1 immunotherapy.

• Language: English
• Keywords: cancer, modelling, immunotherapy, anti-PD-1, iRADIOMICS, non-small-cell lung cancer, MHC class I
• Work type: Doctoral dissertation
• Typology: 2.08 - Doctoral Dissertation
• Organization: FMF - Faculty of Mathematics and Physics
• Year: 2020
• PID: 20.500.12556/RUL-120530
• COBISS.SI-ID: 31081475
• Publication date in RUL: 22.09.2020

Secondary language

• Language: Slovenian
• Title: Fizikalni modeli imunoterapije

Abstract

Imunoterapija z zaviralci imunskih kontrolnih točk je spremenila pristop k zdravljenju raka. Prvič v zgodovini onkologije je mogoče realistično pričakovati popolne ozdravitve nekaterih napredovalih vrst raka, kot sta metastatski melanom in nedrobnocelični rak pljuč, ki sta še do nedavnega veljala za skorajda neozdravljiva. Žal so bolniki s tako dobrimi izidi zdravljenja v manjšini, razlogi pa povečini neznani. Splošni cilj doktorske disertacije je bila prepoznava lastnosti tumorjev, od katerih je odvisen odziv na imunoterapijo s protitelesi receptorja programirane celične smrti 1 (angl. anti-programmed-death-1 (anti-PD-1)), z uporabo računalniškega modeliranja. Za dosego omenjenega cilja smo uporabili dva nasprotna modelska pristopa: mehanistično modeliranje po načelu od spodaj navzgor in podatkovno vodeno modeliranje od zgoraj navzdol. Zaradi zagotovitve zadostne količine eksperimentalnih podatkov, potrebnih za kalibracijo in preverbo modelov, je bilo modeliranje od spodaj navzgor osredotočeno na predkliniko. Razvili smo dva modela, ki smo ju kalibrirali in preverili tako s podatki iz literature kot tudi s podatki lastnih eksperimentov. Oba sta prepoznala pomembno vlogo izražanja poglavitnega kompleksa tkivne skladnosti (angl. major histocompatibility complex (MHC)) razreda I na tumorskih celicah pri odzivu na imunoterapijo anti-PD-1. Nakazala sta, da je mogoče popolne ozdravitve pričakovati le v primeru homogenih tumorjev, kjer vse tumorske celice izražajo MHC razreda I. Podatkovno voden model po načelu od zgoraj navzdol pa smo razvili za potrebe klinične študije, kjer je bila na voljo majhna količina podatkov. Vanjo so bili vključeni bolniki z metastatskim nedrobnoceličnim rakom pljuč, zdravljeni s pembrolizumabom in slikani z [18F]FDG PET/CT. Model je bil osnovan na podlagi radiomskih analiz slik PET primarnih tumorjev (iRADIOMICS) in je pokazal zmožnost napovedovanja bolnikovega celokupnega preživetja. Podrobna analiza zgradbe modela je nakazala, da je boljše odzive na zdravljenje pričakovati v primeru bolj homogenih tumorjev – podobno kot je nakazalo modeliranje od spodaj navzgor. Če bodo ugotovitve iz doktorske disertacije potrjene v neodvisnih študijah, bodo imele pomemben vpliv na klinično prakso, saj bodo omogočile boljše upravljanje bolnikov z rakom, ki se zdravijo z imunoterapijo anti-PD-1.

• Keywords: rak, modeliranje, imunoterapija, anti-PD-1, iRADIOMICS, nedrobnocelični rak pljuč, MHC razreda I