In silico signaling modeling to understand cancer pathways and treatment responses

Abstract

Precision medicine has changed thinking in cancer therapy, highlighting a better understanding of the individual clinical interventions. But what role do the drivers and pathways identified from pan-cancer genome analysis play in the tumor? In this letter, we will highlight the importance of in silico modeling in precision medicine. In the current era of big data, tumor engines and pathways derived from pan-cancer analysis should be integrated into in silico models to understand the mutational tumor status and individual molecular pathway mechanism at a deeper level. This allows to pre-evaluate the potential therapy response and develop optimal patient-tailored treatment strategies which pave the way to support precision medicine in the clinic of the future.

in silico modeling, cancer pathways, precision medicine

Introduction

Precision medicine has changed thinking in cancer therapy and offers new options for better therapy. Mutations in individual cancers accumulate often as so called non-oncogenic ‘passenger’ mutations which are not functionally relevant for cancer progression and thereby also not relevant for the therapeutic outcome of targeted drugs [1]. Others are crucial for oncogenic transformation and are often part of signaling pathways or their close interaction partners. These mutations are promising target candidates and serve in the clinic as biomarker mutations to stratify patient groups by mutation constellations and signatures into those patients who will or will not profit from a tailored targeted therapy. Several common cancer signaling pathways could be identified in large mutation screening efforts across all tumor entities that are suitable for targeted therapeutic intervention [2]. However, though targeted therapies are used in the clinic with initial success, often they are met by emerging resistance [2–4]. Thus, clinically, it is of interest to understand the molecular mechanisms of the resulting clinical interventions based on the patient's mutations and the involved signaling pathways which will help to find the optimal therapy.

Lim et al. [5] systematically evaluated computational tools for pathway analysis based on genomics pan-cancer data sets. But what role do the pathways identified from pan-cancer genome analysis play in the tumor? What are the underlying molecular mechanisms of the pathways regarding tumor growth and treatment response?

Considering the power of systems biological network analysis and simulations, it is important to have comprehensive in silico studies to evaluate individual mutations, the most often encountered constellations and use all the knowledge gained to stratify the patients by these signatures into potential targeted therapy responders and non-responders.

An important part of the new world of precision medicine is thus in silico modeling. It bridges the gap between molecular biology and the clinic with the focus of establishing patient-specific in silico tumor models. Based on sequencing data, individual tumor drivers and related pathway interactions can be integrated into a mathematical network of key signaling cascades which consider systemic responses in tumors such as apoptosis, proliferation and resistance in a mathematical simplified way using signaling proteins (‘nodes’) and their type of action (‘edges’; inhibition and activation) [6, 7].

In silico modeling of cancer pathways requires large and diverse sets of data on functional connections, allowing description of how a signaling network is perturbed by genome alterations [7, 8]. Further data can be included from available literature about mutations or biochemical relevant kinase cascades and patient omics data to correlate this with patient survival (e.g. ClinVar, KEGG). Subsequently, mathematical network modeling allows description of crucial events in cancer progression and systemic therapeutic effects in pathway topologies in silico [6, 9]. For instance, modeling tools such as SQUAD and Jimena interpolate between network states in cancer cells [10, 11], and identify driver nodes, types of network control and regulatory functional states in biological pathways [12]. This enables a comprehensive investigation of the effects of the patient's driver mutations and molecular pathway mechanisms in cancer progression and therapeutic intervention. Subsequently, drug targets and an optimal therapy with a low cancer resistance risk can be calculated in silico [6, 13], tested in relevant in vitro models such as novel tissue-engineered tumor models and then be transferred to the clinic for selected patient cases with a similar mutation constellation [6, 9, 14].

From a clinical perspective, in silico models may also become helpful for a standardized patient stratification and therapy. However, difficulties emerge in the validation process, with suboptimal results in terms of reproducibility of the data in different cohorts of patients or high-risk individuals. For this task, multiparametric methods are essential to improve the capability for early detection or progression of cancer as well as to stratify patients eligible for extensive cancer screening. An iterative cycle of modeling and clinical monitoring and in vitro pre-testing with signaling analysis in different genetic backgrounds allows the validation and refinement of the in silico model [6] but the transfer into a standardized patient stratification and therapy is still a major task to be accomplished.

However, during tumor progression, radiation therapy is often the last resort as sole or additive measure in locally advanced cancer, if functional loss can be avoided. Here, in silico modeling of the tumor can be helpful in determining which pathways should be targeted to achieve radiosensitization.

A key challenge of cancer therapy is the tumor heterogeneity and its influence on the pharmacogenetic drug interaction [7]. On the other hand, the complexity of genomic information influences patient therapy prediction models in precision medicine [15, 16]. Moreover, limitations regarding experimentally validated protein interactions as well as missing pathway annotations reduce the information for in silico modeling which may lead to incorrect predictions. To overcome this, pathway data integration from multiple organisms such as the Pathway Commons web resource is helpful [17]. Further approaches focus on network components. Tumor cells are regulated by hub nodes mainly situated in the central cancer signaling pathways [18]. Identifying the hubs that are essential for tumor progression is a strategy in precision medicine [7].

Several studies demonstrate that integrating multi-omics data supports precision medicine and improves the clinical outcome (reviewed in [7, 16, 19]). For instance, pathway analysis approaches such as Biocarta, Ingenuity, KEGG and PharmaGKB find gene targets and potential drugs by high-throughput screening [7]. Moreover, Kapoor et al. [20] developed a patient-specific therapeutic approach reducing and partly preventing the possibility of recurrence and drug toxicity. Further ‘proof of principle’ approaches focus on drug combinations to explore the signaling network components in which the drug targets are embedded [7, 8, 21]. Such in silico modeling from molecular target to drug action covers a wide range of combination effects, pathway topologies and relationships between targets to find novel targets and optimal patient therapies [8]. Other in silico strategies integrate clinical and genomic data to select drug targets that act as hubs [2, 15, 18]. We developed in silico modeling tools [6, 9, 11, 13, 14] that allow the understanding of tumour-associated pathways and underlying molecular treatment responses. For instance, our recent modeling tool allows patient stratification and prediction of drug effectiveness by integrating key mutations and relevant signaling pathways derived from genomic screening of a tissue-engineered tumor test system [6, 9].

Predictions and suggestions from the models are used in the molecular tumor board of the Comprehensive Cancer Center Mainfranken to support therapy decisions in selected clinical cases. Several studies highlight the incorporation of integrative clinical sequencing data in a multidisciplinary Sequencing Tumor Board to find the optimal therapy for the patient [22, 23]. We exemplify this for patients with advanced cancer (lung cancer, hematological cancer, adrenocortical and oral cancer) that did not respond to therapy. We can point out in silico alternative strategies of targeting pathways considering specific mutational background [6, 9, 14, 24–26] and paving the way for the best targeted therapeutic strategies to overcome resistance in selected patient cases with matching mutation signature.

The aforementioned pathway modeling approaches allow exploration of drug action in underlying cancer-specific pathways including crosstalk signaling interferences. Moreover, they can be adjusted to patient mutational alterations and clinical parameters, allowing the analysis of individual genomic profiles of a patient-derived tumor [6, 7, 20]. This is a step forward in precision medicine, as an optimal targeted mono or combination therapy can be calculated by the combination of biological and medical data with mathematical methods. In this way, insights from in silico modeling complement omics data with relevant targeted pathway implications to support precision medicine [7, 8]. However, for generalization and standardization the selected results still need further validation by patient cohorts and clinical studies which investigate the clinical application of different targeted and combination therapy regimes.

In conclusion, in the current era of big data and increasing number of bioinformatic tools, it is important to have such comprehensive evaluation studies as presented by Lim et al. [5]. Moreover, tumor engines and pathways derived from pan-cancer analysis [2] should be integrated into in silico models to understand the mutational tumor status and individual molecular pathway mechanism at a deeper level. This allows to pre-evaluate the potential therapy response and develop optimal patient-tailored treatment strategies which pave the way to support precision medicine in the clinic.

Funding

Federal Ministry of Education and Research (grant FKZ 031L0129A to GD, FKZ 031L0129B to MK and TD) and Era-Net (grant 01KT1801 to MK and TD).