Boosting predictive models and augmenting patient data with relevant genomic and pathway information.

The recurrence of low-stage lung cancer poses a challenge due to its unpredictable nature and diverse patient responses to treatments. Personalized care and patient outcomes heavily rely on early relapse identification, yet current predictive models, despite their potential, lack comprehensive genetic data. This inadequacy fuels our research focus-integrating specific genetic information, such as pathway scores, into clinical data. Our aim is to refine machine learning models for more precise relapse prediction in early-stage non-small cell lung cancer. To address the scarcity of genetic data, we employ imputation techniques, leveraging publicly available datasets such as The Cancer Genome Atlas (TCGA), integrating pathway scores into our patient cohort from the Cancer Long Survivor Artificial Intelligence Follow-up (CLARIFY) project. Through the integration of imputed pathway scores from the TCGA dataset with clinical data, our approach achieves notable strides in predicting relapse among a held-out test set of 200 patients. By training machine learning models on enriched knowledge graph data, inclusive of triples derived from pathway score imputation, we achieve a promising precision of 82% and specificity of 91%. These outcomes highlight the potential of our models as supplementary tools within tumour, node, and metastasis (TNM) classification systems, offering improved prognostic capabilities for lung cancer patients. In summary, our research underscores the significance of refining machine learning models for relapse prediction in early-stage non-small cell lung cancer. Our approach, centered on imputing pathway scores and integrating them with clinical data, not only enhances predictive performance but also demonstrates the promising role of machine learning in anticipating relapse and ultimately elevating patient outcomes.

A Systems Biology Approach to Understand the Racial Disparities in Colorectal Cancer.

Racial disparities between Black/African Americans (AA) and White patients in colorectal cancer are an ever-growing area of concern. Black/AA show the highest incidence and have the highest mortality among major U.S. racial groups. There is no definite cause other than possible sociodemographic, socioeconomic, education, nutrition, delivery of healthcare, screening, and cultural factors. A primary limitation in this field is the lack of and small sample size of Black/AA studies. Thus, this study aimed to investigate whether differences in gene expression contribute to this ongoing unanswered racial disparity issue. In this study, we examined transcriptomic data of Black/AA and White patient cohorts using a bioinformatic and systems biology approach. We performed a Kaplan-Meier overall survival analysis between both patient cohorts across critical colorectal cancer signal transduction networks (STN), to determine the differences in significant genes across each cohort. Other bioinformatic analyses performed included PROGENy (pathway responsive genes for activity inference), RNA sequencing differential expression using DESeq2, multivariable-adjusted regression, and other associated Kaplan-Meier analyses. These analyses identified novel prognostic genes independent from each cohort, 176 differentially expressed genes, and specific patient cohort STN survival associations. Despite the overarching limitation, the results revealed several novel differences in gene expression between the colorectal cancer Black/AA and White patient cohorts, which allows one to dive deeper into and understand the behavior on a systems level of what could be driving this racial difference across colorectal cancer. Concretely, this information can guide precision medicine approaches tailored specifically for colorectal cancer racial disparities. SIGNIFICANCE: The purpose of this work is to investigate the racial disparities in colorectal cancer between Black/AA and White patient cohorts using a systems biology and bioinformatic approach. Our study investigates the underlying biology of each patient cohort. Concretely, the findings of this study include disparity-associated genes and pathways, which provide a tangible starting point to guide precision medicine approaches tailored specifically for colorectal cancer racial disparities.

Synergy between imputed genetic pathway and clinical information for predicting recurrence in early stage non-small cell lung cancer.

OBJECTIVE: Lung cancer exhibits unpredictable recurrence in low-stage tumors and variable responses to different therapeutic interventions. Predicting relapse in early-stage lung cancer can facilitate precision medicine and improve patient survivability. While existing machine learning models rely on clinical data, incorporating genomic information could enhance their efficiency. This study aims to impute and integrate specific types of genomic data with clinical data to improve the accuracy of machine learning models for predicting relapse in early-stage, non-small cell lung cancer patients. METHODS: The study utilized a publicly available TCGA lung cancer cohort and imputed genetic pathway scores into the Spanish Lung Cancer Group (SLCG) data, specifically in 1348 early-stage patients. Initially, tumor recurrence was predicted without imputed pathway scores. Subsequently, the SLCG data were augmented with pathway scores imputed from TCGA. The integrative approach aimed to enhance relapse risk prediction performance. RESULTS: The integrative approach achieved improved relapse risk prediction with the following evaluation metrics: an area under the precision-recall curve (PR-AUC) score of 0.75, an area under the ROC (ROC-AUC) score of 0.80, an F1 score of 0.61, and a Precision of 0.80. The prediction explanation model SHAP (SHapley Additive exPlanations) was employed to explain the machine learning model's predictions. CONCLUSION: We conclude that our explainable predictive model is a promising tool for oncologists that addresses an unmet clinical need of post-treatment patient stratification based on the relapse risk while also improving the predictive power by incorporating proxy genomic data not available for specific patients.

Memory of stochastic single-cell apoptotic signaling promotes chemoresistance in neuroblastoma.

Gene expression noise is known to promote stochastic drug resistance through the elevated expression of individual genes in rare cancer cells. However, we now demonstrate that chemoresistant neuroblastoma cells emerge at a much higher frequency when the influence of noise is integrated across multiple components of an apoptotic signaling network. Using a JNK activity biosensor with longitudinal high-content and in vivo intravital imaging, we identify a population of stochastic, JNK-impaired, chemoresistant cells that exist because of noise within this signaling network. Furthermore, we reveal that the memory of this initially random state is retained following chemotherapy treatment across a series of in vitro, in vivo, and patient models. Using matched PDX models established at diagnosis and relapse from individual patients, we show that HDAC inhibitor priming cannot erase the memory of this resistant state within relapsed neuroblastomas but improves response in the first-line setting by restoring drug-induced JNK activity within the chemoresistant population of treatment-naïve tumors.

A Systems Biology Approach to Investigate Kinase Signal Transduction Networks That Are Involved in Triple Negative Breast Cancer Resistance to Cisplatin.

Triple negative breast cancer (TNBC) remains a therapeutic challenge due to the lack of targetable genetic alterations and the frequent development of resistance to the standard cisplatin-based chemotherapies. Here, we have taken a systems biology approach to investigate kinase signal transduction networks that are involved in TNBC resistance to cisplatin. Treating a panel of cisplatin-sensitive and cisplatin-resistant TNBC cell lines with a panel of kinase inhibitors allowed us to reconstruct two kinase signalling networks that characterise sensitive and resistant cells. The analysis of these networks suggested that the activation of the PI3K/AKT signalling pathway is critical for cisplatin resistance. Experimental validation of the computational model predictions confirmed that TNBC cell lines with activated PI3K/AKT signalling are sensitive to combinations of cisplatin and PI3K/AKT pathway inhibitors. Thus, our results reveal a new therapeutic approach that is based on identifying targeted therapies that synergise with conventional chemotherapies.

Integration of Clinical Information and Imputed Aneuploidy Scores to Enhance Relapse Prediction in Early Stage Lung Cancer Patients.

Early-stage lung cancer is crucial clinically due to its insidious nature and rapid progression. Most of the prediction models designed to predict tumour recurrence in the early stage of lung cancer rely on the clinical or medical history of the patient. However, their performance could likely be improved if the input patient data contained genomic information. Unfortunately, such data is not always collected. This is the main motivation of our work, in which we have imputed and integrated specific type of genomic data with clinical data to increase the accuracy of machine learning models for prediction of relapse in early-stage, non-small cell lung cancer patients. Using a publicly available TCGA lung adenocarcinoma cohort of 501 patients, their aneuploidy scores were imputed into similar records in the Spanish Lung Cancer Group (SLCG) data, more specifically a cohort of 1348 early-stage patients. First, the tumor recurrence in those patients was predicted without the imputed aneuploidy scores. Then, the SLCG data were enriched with the aneuploidy scores imputed from TCGA. This integrative approach improved the prediction of the relapse risk, achieving area under the precision-recall curve (PR-AUC) score of 0.74, and area under the ROC (ROC-AUC) score of 0.79. Using the prediction explanation model SHAP (SHapley Additive exPlanations), we further explained the predictions performed by the machine learning model. We conclude that our explainable predictive model is a promising tool for oncologists that addresses an unmet clinical need of post-treatment patient stratification based on the relapse risk, while also improving the predictive power by incorporating proxy genomic data not available for the actual specific patients.

Relationship Between Dimensionality and Convergence of Optimization Algorithms: A Comparison Between Data-Driven Normalization and Scaling Factor-Based Methods Using PEPSSBI.

Ordinary differential equation models are used to represent intracellular signaling pathways in silico, aiding and guiding biological experiments to elucidate intracellular regulation. To construct such quantitative and predictive models of intracellular interactions, unknown parameters need to be estimated. Most of biological data are expressed in relative or arbitrary units, raising the question of how to compare model simulations with data. It has recently been shown that for models with large number of unknown parameters, fitting algorithms using a data-driven normalization of the simulations (DNS) performs best in terms of the convergence time and parameter identifiability. DNS approach compares model simulations and corresponding data both normalized by the same normalization procedure, without requiring additional parameters to be estimated, as necessary for widely used scaling factor-based methods. However, currently there is no parameter estimation software that directly supports DNS. In this chapter, we show how to apply DNS to dynamic models of systems and synthetic biology using PEPSSBI (Parameter Estimation Pipeline for Systems and Synthetic Biology). PEPSSBI is the first software that supports DNS, through algorithmically supported data normalization and objective function construction. PEPSSBI also supports model import using SBML and repeated parameter estimation runs executed in parallel either on a personal computer or a multi-CPU cluster.

Systems Biology in ELIXIR: modelling in the spotlight.

In this white paper, we describe the founding of a new ELIXIR Community - the Systems Biology Community - and its proposed future contributions to both ELIXIR and the broader community of systems biologists in Europe and worldwide. The Community believes that the infrastructure aspects of systems biology - databases, (modelling) tools and standards development, as well as training and access to cloud infrastructure - are not only appropriate components of the ELIXIR infrastructure, but will prove key components of ELIXIR's future support of advanced biological applications and personalised medicine. By way of a series of meetings, the Community identified seven key areas for its future activities, reflecting both future needs and previous and current activities within ELIXIR Platforms and Communities. These are: overcoming barriers to the wider uptake of systems biology; linking new and existing data to systems biology models; interoperability of systems biology resources; further development and embedding of systems medicine; provisioning of modelling as a service; building and coordinating capacity building and training resources; and supporting industrial embedding of systems biology. A set of objectives for the Community has been identified under four main headline areas: Standardisation and Interoperability, Technology, Capacity Building and Training, and Industrial Embedding. These are grouped into short-term (3-year), mid-term (6-year) and long-term (10-year) objectives.

Personalised Medicine for Colorectal Cancer Using Mechanism-Based Machine Learning Models.

Gaining insight into the mechanisms of signal transduction networks (STNs) by using critical features from patient-specific mathematical models can improve patient stratification and help to identify potential drug targets. To achieve this, these models should focus on the critical STNs for each cancer, include prognostic genes and proteins, and correctly predict patient-specific differences in STN activity. Focussing on colorectal cancer and the WNT STN, we used mechanism-based machine learning models to identify genes and proteins with significant associations to event-free patient survival and predictive power for explaining patient-specific differences of STN activity. First, we identified the WNT pathway as the most significant pathway associated with event-free survival. Second, we built linear-regression models that incorporated both genes and proteins from established mechanistic models in the literature and novel genes with significant associations to event-free patient survival. Data from The Cancer Genome Atlas and Clinical Proteomic Tumour Analysis Consortium were used, and patient-specific STN activity scores were computed using PROGENy. Three linear regression models were built, based on; (1) the gene-set of a state-of-the-art mechanistic model in the literature, (2) novel genes identified, and (3) novel proteins identified. The novel genes and proteins were genes and proteins of the extant WNT pathway whose expression was significantly associated with event-free survival. The results show that the predictive power of a model that incorporated novel event-free associated genes is better compared to a model focussing on the genes of a current state-of-the-art mechanistic model. Several significant genes that should be integrated into future mechanistic models of the WNT pathway are DVL3, FZD5, RAC1, ROCK2, GSK3B, CTB2, CBT1, and PRKCA. Thus, the study demonstrates that using mechanistic information in combination with machine learning can identify novel features (genes and proteins) that are important for explaining the STN heterogeneity between patients and their association to clinical outcomes.

Personalized Medicine for Neuroblastoma: Moving from Static Genotypes to Dynamic Simulations of Drug Response.

High-risk neuroblastoma is an aggressive childhood cancer that is characterized by high rates of chemoresistance and frequent metastatic relapse. A number of studies have characterized the genetic and epigenetic landscape of neuroblastoma, but due to a generally low mutational burden and paucity of actionable mutations, there are few options for applying a comprehensive personalized medicine approach through the use of targeted therapies. Therefore, the use of multi-agent chemotherapy remains the current standard of care for neuroblastoma, which also conceptually limits the opportunities for developing an effective and widely applicable personalized medicine approach for this disease. However, in this review we outline potential approaches for tailoring the use of chemotherapy agents to the specific molecular characteristics of individual tumours by performing patient-specific simulations of drug-induced apoptotic signalling. By incorporating multiple layers of information about tumour-specific aberrations, including expression as well as mutation data, these models have the potential to rationalize the selection of chemotherapeutics contained within multi-agent treatment regimens and ensure the optimum response is achieved for each individual patient.

BAX and SMAC regulate bistable properties of the apoptotic caspase system.

The initiation of apoptosis is a core mechanism in cellular biology by which organisms control the removal of damaged or unnecessary cells. The irreversible activation of caspases is essential for apoptosis, and mathematical models have demonstrated that the process is tightly regulated by positive feedback and a bistable switch. BAX and SMAC are often dysregulated in diseases such as cancer or neurodegeneration and are two key regulators that interact with the caspase system generating the apoptotic switch. Here we present a mathematical model of how BAX and SMAC control the apoptotic switch. Formulated as a system of ordinary differential equations, the model summarises experimental and computational evidence from the literature and incorporates the biochemical mechanisms of how BAX and SMAC interact with the components of the caspase system. Using simulations and bifurcation analysis, we find that both BAX and SMAC regulate the time-delay and activation threshold of the apoptotic switch. Interestingly, the model predicted that BAX (not SMAC) controls the amplitude of the apoptotic switch. Cell culture experiments using siRNA mediated BAX and SMAC knockdowns validated this model prediction. We further validated the model using data of the NCI-60 cell line panel using BAX protein expression as a cell-line specific parameter and show that model simulations correlated with the cellular response to DNA damaging drugs and established a defined threshold for caspase activation that could distinguish between sensitive and resistant melanoma cells. In summary, we present an experimentally validated dynamic model that summarises our current knowledge of how BAX and SMAC regulate the bistable properties of irreversible caspase activation during apoptosis.

Accurate prediction of kinase-substrate networks using knowledge graphs.

Phosphorylation of specific substrates by protein kinases is a key control mechanism for vital cell-fate decisions and other cellular processes. However, discovering specific kinase-substrate relationships is time-consuming and often rather serendipitous. Computational predictions alleviate these challenges, but the current approaches suffer from limitations like restricted kinome coverage and inaccuracy. They also typically utilise only local features without reflecting broader interaction context. To address these limitations, we have developed an alternative predictive model. It uses statistical relational learning on top of phosphorylation networks interpreted as knowledge graphs, a simple yet robust model for representing networked knowledge. Compared to a representative selection of six existing systems, our model has the highest kinome coverage and produces biologically valid high-confidence predictions not possible with the other tools. Specifically, we have experimentally validated predictions of previously unknown phosphorylations by the LATS1, AKT1, PKA and MST2 kinases in human. Thus, our tool is useful for focusing phosphoproteomic experiments, and facilitates the discovery of new phosphorylation reactions. Our model can be accessed publicly via an easy-to-use web interface (LinkPhinder).

Applications of personalised signalling network models in precision oncology.

As our ability to provide in-depth, patient-specific characterisation of the molecular alterations within tumours rapidly improves, it is becoming apparent that new approaches will be required to leverage the power of this data and derive the full benefit for each individual patient. Systems biology approaches are beginning to emerge within this field as a potential method of incorporating large volumes of network level data and distilling a coherent, clinically-relevant prediction of drug response. However, the initial promise of this developing field is yet to be realised. Here we argue that in order to develop these precise models of individual drug response and tailor treatment accordingly, we will need to develop mathematical models capable of capturing both the dynamic nature of drug-response signalling networks and key patient-specific information such as mutation status or expression profiles. We also review the modelling approaches commonly utilised within this field, and outline recent examples of their use in furthering the application of systems biology for a precision medicine approach to cancer treatment.

Understanding the role of distance, direction and cue salience in an associative model of landmark learning.

Navigation and spatial memory relies on the ability to use and recall environmental landmarks relative to important locations. Such learning is thought to result from the strengthening of associations between the goal location and environmental cues. Factors that contribute to the strength of this association include cue stability, saliency and cue location. Here we combine an autoregressive random walk model, that describes goal-directed swimming behaviour, with an associative learning model to provide an integrated model of landmark learning, using the water maze task. The model allows for the contribution of each cue, the salience and the vector information provided (both distance and directional) to be separately analysed. The model suggests that direction and distance information are independent components and can influence searching patterns. Importantly, the model can also be used to simulate various experimental scenarios to understand what has been learnt in relation to the cues, thereby offering new insights into how animals navigate.

The extracellular matrix as a key regulator of intracellular signalling networks.

The extracellular matrix (ECM) is a salient feature of all solid tissues within the body. This complex, acellular entity is composed of hundreds of individual molecules whose assembly, architecture and biomechanical properties are critical to controlling the behaviour and phenotype of the different cell types residing within tissues. Cells are the basic unit of life and the core building block of tissues and organs. At their simplest, they follow a set of rules, governed by their genetic code and effected through the complex protein signalling networks that these genes encode. These signalling networks assimilate and process the information received by the cell to control cellular decisions that govern cell fate. The ECM is the biggest provider of external stimuli to cells and as such is responsible for influencing intracellular signalling dynamics. In this review, we discuss the inclusion of ECM as a central regulatory signalling sub-network in computational models of cellular decision making, with a focus on its role in diseases such as cancer. LINKED ARTICLES: This article is part of a themed section on Translating the Matrix. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v176.1/issuetoc.

Impact of measurement noise, experimental design, and estimation methods on Modular Response Analysis based network reconstruction.

Modular Response Analysis (MRA) is a method to reconstruct signalling networks from steady-state perturbation data which has frequently been used in different settings. Since these data are usually noisy due to multi-step measurement procedures and biological variability, it is important to investigate the effect of this noise onto network reconstruction. Here we present a systematic study to investigate propagation of noise from concentration measurements to network structures. Therefore, we design an in silico study of the MAPK and the p53 signalling pathways with realistic noise settings. We make use of statistical concepts and measures to evaluate accuracy and precision of individual inferred interactions and resulting network structures. Our results allow to derive clear recommendations to optimize the performance of MRA based network reconstruction: First, large perturbations are favorable in terms of accuracy even for models with non-linear steady-state response curves. Second, a single control measurement for different perturbation experiments seems to be sufficient for network reconstruction, and third, we recommend to execute the MRA workflow with the mean of different replicates for concentration measurements rather than using computationally more involved regression strategies.

Systems biology: old news or new stimulus for biochemistry.

In this issue of Essays in Biochemistry, biochemistry meets systems biology-a blind date that may hold all the promises, pitfalls and failures of a relationship where a new discipline has been sprung upon a well-established one. As the articles in this issue show, the blind date in this case has great potential to develop into a long-term relationship, where both partners share common values but can benefit from different complementary approaches. Together this partnership is well poised to address and solve some of the major challenges in modern biology.

MASTL overexpression promotes chromosome instability and metastasis in breast cancer.

MASTL kinase is essential for correct progression through mitosis, with loss of MASTL causing chromosome segregation errors, mitotic collapse and failure of cytokinesis. However, in cancer MASTL is most commonly amplified and overexpressed. This correlates with increased chromosome instability in breast cancer and poor patient survival in breast, ovarian and lung cancer. Global phosphoproteomic analysis of immortalised breast MCF10A cells engineered to overexpressed MASTL revealed disruption to desmosomes, actin cytoskeleton, PI3K/AKT/mTOR and p38 stress kinase signalling pathways. Notably, these pathways were also disrupted in patient samples that overexpress MASTL. In MCF10A cells, these alterations corresponded with a loss of contact inhibition and partial epithelial-mesenchymal transition, which disrupted migration and allowed cells to proliferate uncontrollably in 3D culture. Furthermore, MASTL overexpression increased aberrant mitotic divisions resulting in increased micronuclei formation. Mathematical modelling indicated that this delay was due to continued inhibition of PP2A-B55, which delayed timely mitotic exit. This corresponded with an increase in DNA damage and delayed transit through interphase. There were no significant alterations to replication kinetics upon MASTL overexpression, however, inhibition of p38 kinase rescued the interphase delay, suggesting the delay was a G2 DNA damage checkpoint response. Importantly, knockdown of MASTL, reduced cell proliferation, prevented invasion and metastasis of MDA-MB-231 breast cancer cells both in vitro and in vivo, indicating the potential of future therapies that target MASTL. Taken together, these results suggest that MASTL overexpression contributes to chromosome instability and metastasis, thereby decreasing breast cancer patient survival.

Personalized Computational Models as Biomarkers.

Biomarkers are cornerstones of clinical medicine, and personalized medicine, in particular, is highly dependent on reliable and highly accurate biomarkers for individualized diagnosis and treatment choice. Modern omics technologies, such as genome sequencing, allow molecular profiling of individual patients with unprecedented resolution, but biomarkers based on these technologies often lack the dynamic element to follow the progression of a disease or response to therapy. Here, we discuss computational models as a new conceptual approach to biomarker discovery and design. Being able to integrate a large amount of information, including dynamic information, computational models can simulate disease evolution and response to therapy with high sensitivity and specificity. By populating these models with personal data, they can be highly individualized and will provide a powerful new tool in the armory of personalized medicine.

Performance of objective functions and optimisation procedures for parameter estimation in system biology models.

Mathematical modelling of signalling pathways aids experimental investigation in system and synthetic biology. Ever increasing data availability prompts the development of large dynamic models with numerous parameters. In this paper, we investigate how the number of unknown parameters affects the convergence of three frequently used optimisation algorithms and four objective functions. We compare objective functions that use data-driven normalisation of the simulations with those that use scaling factors. The data-driven normalisation of the simulation approach implies that simulations are normalised in the same way as the data, making both directly comparable. The scaling factor approach, which is commonly used for parameter estimation in dynamic systems, introduces scaling factors that multiply the simulations to convert them to the scale of the data. Here we show that the scaling factor approach increases, compared to data-driven normalisation of the simulations, the degree of practical non-identifiability, defined as the number of directions in the parameter space, along which parameters are not identifiable. Further, the results indicate that data-driven normalisation of the simulations greatly improve the speed of convergence of all tested algorithms when the overall number of unknown parameters is relatively large (74 parameters in our test problems). Data-driven normalisation of the simulations also markedly improve the performance of the non-gradient-based algorithm tested even when the number of unknown parameters is relatively small (10 parameters in our test problems). As the models and the unknown parameters increase in size, the data-driven normalisation of the simulation approach can be the preferred option, because it does not aggravate non-identifiability and allows for obtaining parameter estimates in a reasonable amount of time.