No user image
Anonymous Feb 03
Bioinformatics Modelling for Immunotherapy image

Bioinformatics Modelling for Immunotherapy

Precision medicine requires the delivery of individually adapted medical care based on the genetic characteristics of each patient and their tumour. Today, technological advances such as next-generation sequencing are providing unprecedented opportunities to draw a comprehensive picture of the tumour genomics landscape and ultimately enable individualised treatment

Several precision medicine clinical trials are underway using treatment algorithms to assign patients to specific targeted therapies based on tumour molecular. Examples of these trials include metastatic disease from all cancer types such as the SHIVA trial, MPACT and the WINTHER trial, as well as disease-specific trials such as the SAFIRO2 trials.


Some of the criteria taken into account to design treatment algorithms for future precision medicine trials include specification of the technology used for molecular profiling, the definition of ‘targetable’ molecular alterations and targeted agents and the prioritisation of targetable molecular alterations in patients whose tumours have more than a single alteration. It goes without saying that in order to evaluate the efficiency of treatment algorithms in guiding therapy, it needs to produce interpretable and reproducible results [Le Tourneau et al, 2015].


A step beyond this is the added complexity of the tumour-immune cell interactions and the impact of immune-oncology agents on this. These agents are particularly exciting because they can induce a durable anti-tumour response, with some patients achieving disease control for many years. However, not all patients benefit and researchers are now investing their efforts into finding predictive biomarkers of response and resistance. For computational biology, our lack of understanding of tumour biology and drug pharmacology makes treatment algorithms difficult to design [Harris, et al 2016].


The success of cancer immunotherapies today is based on agents that induce or augment immune responses to cancer. This takes the form of monotherapies – use of checkpoint blockers, vaccination with neoantigens and adoptive T cell transfer. Additionally, we have combinations of these therapies as well as their combinations of targeted therapies.  An infrastructure for precision oncology based on these strategies requires state-of-the-art molecular tools, such as the NGS instruments and dedicated IT systems depicted in the image below [Hackl, et al 2016]:


Modeling tumour-immune interactions

In order to model tumour-immune system interactions (both adaptive and innate immune responses), researchers may focus on the dynamics of each population of interest (e.g. Lymphocytes or tumour cells) with a differential equation reflecting changes to the population over time. 15 or 20 coupled differential equations may be used for each cellular subtype or chemical mediator concentration to describe immune responses. However, there is a tendency for modellers to utilise more simple, parameterizable models as a trade-off between complexity and computational cost.


A general and concise example of the structure of tumour-immune interaction equations can be given by the following pair of ordinary differential equations (ODEs): 

In the example above, populations T and I represent tumour and immune cell populations, respectively. Function f(T) represents net tumour growth; exponential, logistic and Gompertz functions are all commonly used (you can find further detail on these models as well as several alternatives in ‘Classical mathematical models for description and prediction of experimental tumour growth’ [Benzekry et al 2014]). Cytotoxic effector lymphocytes are commonly modelled as the immune population in the above simple but robust system of ODEs. A simple schematic of these basic tumour-immune population dynamics can be seen here:

This fundamental structure representing simple lymphocyte-tumour interactions has been extensively adapted to incorporate more complex features of the tumour-immune relationship over the last several decades, including cellular heterogeneity, spatial dynamics, delayed feedback, and cytokine activity and other signalling and modulating factors.


Despite the key role in the immune response of cytotoxic effector cells, insights into the behaviour of other immune cell populations can also be gained from simple mathematical models. For example, simulations of interactions between macrophages and T lymphocytes accounting for antigen presentation have demonstrated that the magnitude of the adaptive response can be dramatically increased by the early activation of CD4+ helper T cells, which allows early macrophage accumulation of tumour cell debris and a rapidly mounting immune attack. Regulatory T cells can also influence tumour development and dormancy; in tumours featuring low immunogenicity and a high growth rate, effector T cells are able to outcompete regulatory T cells but are still unable to control the tumour.


These simple ODE models are useful in investigations towards temporal population dynamics but to thoroughly analyse certain elements of the immune response, spatial variation must be taken into account. For example, heterogeneous spatial distributions of tumour and immune cell populations have been found in immunogenic tumours [Matazavinos & Chaplain, 2004]. Additionally, the effect of cytokines, which mediate tumour-immune interactions, add a further level of model complexity in which reaction or diffusion processes such as the exchange of substances with the local microenvironment may need to be accounted for.


Modeling treatment outcomes

Making the transition from concept to clinic in the case of immunotherapies can be a long and challenging process, and few protocols have been established to guide standard treatment. Even with the exciting rate of FDA approvals in the immune-oncology space, our understanding of the field lags far behind. To overcome this there has been increased utilization of mathematical models designed to predict the outcomes of hypothetical treatment scenarios to offer insights into the mechanisms underlying the potential success or failure of a given therapy.


The most widely studied area is the interleukin family. IL-2, primarily responsible for the activation, growth and differentiation of lymphocytes, is one of few cytokine-based treatments to receive FDA approval. When applied in isolation, models have observed the potential for an immune response capable of growing without bounds which could lead to harmful side effects such as capillary leak syndrome in clinical patients [Kirschner & Panetta 1997], although if both tumor antigenicity and the level of IL-2 are high, this side effect may be reduced [Banerjee 2008].


To improve specificity and accuracy in modelling for use in clinical practice much more temporally and spatially resolved clinical data are desperately needed. The availability of more clinical data will allow investigators to develop more accurate mathematical models of specific cancers and treatments by incorporating more biologically realistic, detailed characteristics of tumour-immune interactions [Walker & Enderling 2015].