Computational and systems oncology最新文献

Cancer cell plasticity is the ability of tumor cells to switch phenotypes and is one of the predominant requisites of cancer cells capable of undergoing metastasis. Cancer cell plasticity is also recognized as one of the major contributors to intratumoral heterogeneity, a critical factor underlying the progression of malignant tumors, which is known to modify tumor response and induce resistance against various modes of therapy, thus posing a barrier to efficient cancer management. Cancer cell plasticity is acquired by the subversion of cell signaling pathways like mitogen-activated protein kinase pathway, phosphoinositide-3-kinase, signal transducer and activator of transcription 3, Wnt, Hedgehog and Notch as well as cellular programs such as epithelial to mesenchymal transition and phenotypic plasticity. This complex phenomenon has been studied in many cancer types like pancreatic cancer, colon cancer and breast cancer. This review will explore the current understanding we have in breast cancer on the intrinsic molecular mechanisms of cancer cell plasticity and the resistance to various types of cancer therapy that arise as a result of plasticity. We conclude by exploring the potential novel therapies that specifically target the pathways leading to plasticity and can be leveraged to treat patients living with the disease.

The remarkable flexibility and adaptability of generative adversarial networks (GANs) have led to the proliferation of its models in bioinformatics research. Proteomic and transcriptomic profiles have been shown to be promising methods for discovering and identifying disease biomarkers. However, those analyses were performed by trained human examiners making the process tedious, time consuming, and hard to standardize. With the development of GANs, it is now possible to reduce computational costs and human time for bioinformatics analysis to produce effective biomarkers. Moreover, GANs help address the lack of phenotypic state transitional gene expression data as well as avoid protected human data constraints by generating RNA sequencing (RNA-seq) data from random vectors. The purpose of this review is to summarize the use of GAN approaches and techniques to augment RNA-seq expression data and identify clinically useful biomarkers. We compare different studies that use different types of GAN models to examine the biomarkers. Also, we identify research gaps and challenges that apply GANs to bio-informatics. Finally, we propose potential directions for future research.

Biological systems across various length and time scales are noisy, including tissues. Why are biological tissues inherently chaotic? Does heterogeneity play a role in determining the physiology and pathology of tissues? How do physical and biochemical heterogeneity crosstalk to dictate tissue function? In this review, we begin with a brief primer on heterogeneity in biological tissues. Then, we take examples from recent literature indicating functional relevance of biochemical and physical heterogeneity and discuss the impact of heterogeneity on tissue function and pathology. We take specific examples from studies on epithelial tissues to discuss the potential role of inherent tissue heterogeneity in tumorigenesis.

Cancer is a life-threatening process that stems from genetic mutations in cells, which leads to the formation of tumors, and is a major cause of deaths in the United States, with secondary metastasis being a major driver of fatality. The development of an optimal metastatic environment is an essential process prior to tumor metastasis. This process, called pre-metastatic niche formation, involves the activation of resident fibroblast-like cells and macrophages. Tumor-mediated factors introduced to this environment transform resident cells that secrete additional growth factors and remodel the extracellular matrix, which is thought to promote tumor colonization and metastasis in the secondary environment. Furthermore, an important component of metastasis is the biological process of epithelial–mesenchymal transition, which can be exploited by cancer cells to change their phenotype, to migrate and proliferate as necessary. In this review, we discuss recent advances in the investigation of cancer growth and migration. Computational models that focus on biochemical signaling and multicellular dynamics are examined. Machine learning models and image analysis that classify cancer-related data are also explored. Through this review, we highlight advances in the study of important aspects of cancer and metastasis signaling and computational tools to study these dynamics.

Misexpression and remodeling of the extracellular matrix is a canonical hallmark of cancer, although the extent of cancer-associated aberrations in the genes coding for extracellular matrix (ECM) proteins and the consequences thereof are not well understood. In this study, we examined the alterations in core matrisomal genes across a set of nine cancers. These genes, especially the ones encoding for ECM glycoproteins (GP), were observed to be more susceptible to mutations than copy number variations across cancers. We classified the glycoprotein genes based on the ubiquity of their mutations across the nine cancer groups and estimated their evolutionary age using phylostratigraphy. To our surprise, the ECM glycoprotein genes commonly mutated across all cancers were predominantly unicellular in origin, whereas those commonly showing mutations in specific cancers evolved mostly during and after the unicellular-multicellular transition. Pathway annotation for biological interactions revealed that the most pervasively mutated glycoprotein set regulated a larger set of inter-protein interactions and constituted more cohesive interaction networks relative to the cancer-specific mutated set. In addition, ontological prediction revealed the pervasively mutated set to be strongly enriched for basement membrane (BM) dynamics. Our results suggest that ancient unicellular-origin ECM GP were canalized into playing critical tissue morphogenetic roles, and when disrupted through matrisomal gene mutations, associated with neoplastic transformation of a wide set of human tissues.

Oxygenation of tumors and the effect of hypoxia on cancer cell metabolism is a widely studied subject. Hypoxia-inducible factor (HIF), the main actor in the cell response to hypoxia, represents a potential target in cancer therapy. HIF is involved in many biological processes such as cell proliferation, survival, apoptosis, angiogenesis, iron metabolism, and glucose metabolism. This protein regulates the expressions of lactate dehydrogenase (LDH) and pyruvate dehydrogenase (PDH), both essential for the conversion of pyruvate to be used in aerobic and anaerobic pathways. HIF upregulates LDH, increasing the conversion of pyruvate into lactate which leads to higher secretion of lactic acid by the cell and reduced pH in the microenvironment. HIF indirectly downregulates PDH, decreasing the conversion of pyruvate into acetyl coenzyme A, which leads to reduced usage of the tricarboxylic acid (TCA) cycle in aerobic pathways. Upregulation of HIF may promote the use of anaerobic pathways for energy production even in normal extracellular oxygen conditions. Higher use of glycolysis even in normal oxygen conditions is called the Warburg effect. In this paper, we focus on HIF variations during tumor growth and study, through a mathematical model, its impact on the two metabolic key genes PDH and LDH, to investigate its role in the emergence of the Warburg effect. Mathematical equations describing the enzyme regulation pathways were solved for each cell of the tumor represented in an agent-based model to best capture the spatio-temporal oxygen variations during tumor development caused by cell consumption and reduced diffusion inside the tumor. Simulation results show that reduced HIF degradation in normoxia can induce higher lactic acid production. The emergence of the Warburg effect appears after the first period of hypoxia before oxygen conditions return to a normal level. The results also show that targeting the upregulation of LDH and the downregulation of PDH could be relevant in therapy.

In the tumor microenvironment (TME), functional interactions among tumor, immune, and stromal cells and the extracellular matrix play key roles in tumor progression, invasion, immune modulation, and response to treatment. Intra-tumor heterogeneity is ubiquitous not only at the genetic and transcriptomic levels but also in the composition and characteristics of TME. However, quantitative inference on spatial heterogeneity in the TME is still limited. Here, we propose a framework to use network graph-based spatial statistical models on spatially annotated molecular data to gain insights into modularity and spatial heterogeneity in the TME. Applying the framework to spatial transcriptomics data from pancreatic ductal adenocarcinoma samples, we observed significant global and local spatially correlated patterns in the abundance score of tumor cells; in contrast, immune cell types showed dispersed patterns in the TME. Hypoxia, EMT, and inflammation signatures contributed to intra-tumor spatial variations. Spatial patterns in cell type abundance and pathway signatures in the TME potentially impact tumor growth dynamics and cancer hallmarks. Tumor biopsies are integral to the diagnosis and clinical management of cancer patients; our data suggest that owing to intra-tumor non-genetic spatial heterogeneity, individual biopsies may underappreciate the extent of clinically relevant, functional variations across geographic regions within tumors.

In vitro migration assays are a cornerstone of cell biology and have found extensive utility in research. Over the past decade, several variations of the two-dimensional (2D) migration assay have improved our understanding of this fundamental process. However, the ability of these approaches to capture the functional heterogeneity during migration and their accessibility to inexperienced users has been limited. We downloaded published time-lapse 2D cell migration data sets and subjected them to feature extraction with the Fiji software. We used the “Analyze Particles” tool to extract 10 cell geometry features (CGFs), which were grouped into “shape,” “size,” and “position” descriptors. Next, we defined the migratory status of cells using the “MTrack2” plugin. All data obtained from Fiji were further subjected to rigorous statistical analysis with R version 4.0.2. We observed consistent associative trends between size and shape descriptors and validated our observations across four independent data sets. We used these descriptors to identify and characterize “nonmigrator (NM)” and “migrator (M)” subsets. Statistical analysis allowed us to identify considerable heterogeneity in the NM subset. Interestingly, differences in 2D-packing appeared to affect CGF trends and heterogeneity within the migratory subsets. We developed an analytical pipeline using open source tools, to identify and morphologically characterize functional migratory subsets from label-free, time-lapse imaging data. Our quantitative approach identified heterogeneity between nonmigratory cells and predicted the influence of 2D-packing on migration.

Ovarian cancer is commonly diagnosed in its late stages, and new treatment modalities are needed to improve patient outcomes and survival. We have recently established the synergistic effects of combination tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) and procaspase activating compound (PAC-1) therapies in granulosa cell tumors (GCT) of the ovary, a rare form of ovarian cancer, using a mathematical model of the effects of both drugs in a GCT cell line. Here, to understand the mechanisms of combined TRAIL and PAC-1 therapy, study the viability of this treatment strategy, and accelerate preclinical translation, we leveraged our mathematical model in combination with population pharmacokinetics (PKs) models of both TRAIL and PAC-1 to expand a realistic heterogeneous cohort of virtual patients and optimize treatment schedules. Using this approach, we investigated treatment responses in this virtual cohort and determined optimal therapeutic schedules based on patient-specific PK characteristics. Our results showed that schedules with high initial doses of PAC-1 were required for therapeutic efficacy. Further analysis of individualized regimens revealed two distinct groups of virtual patients within our cohort: one with high PAC-1 elimination and one with normal PAC-1 elimination. In the high elimination group, high weekly doses of both PAC-1 and TRAIL were necessary for therapeutic efficacy; however, virtual patients in this group were predicted to have a worse prognosis when compared to those in the normal elimination group. Thus, PAC-1 PK characteristics, particularly clearance, can be used to identify patients most likely to respond to combined PAC-1 and TRAIL therapy. This work underlines the importance of quantitative approaches in preclinical oncology.

Intratumor heterogeneity (ITH) is an omnipresent property of cancers and predicts poor survival in most types of cancer. The propensity to metastasize and the regrowth of tumors after therapy are both associated with ITH. Quantification of the level of ITH in a malignancy is hence of great interest, and accurate inference of ITH could guide clinical decision making. However, ITH is an emergent property of billions of cells and requires mathematical modeling for inference from a limited number of measurements. Over the last decade, numerous mathematical and computational models have been introduced to infer ITH from variant-allele frequencies, copy number variations, or from data of experimental model systems. These quantitative modeling efforts have advanced the understanding of tumor evolution, underlined poor prognosis associated with ITH, and elucidated the importance of functional heterogeneity, that is, cancer stem cells. Yet, a comprehensive overview of the different mathematical models, their underlying assumptions, their limitations, and their strengths is missing. In this Perspective, we highlight the achievements of mathematical modeling and present a framework which allows better understanding of the mathematical models themselves.