Leveraging Tissue-Specific Enhancer-Target Gene Regulatory Networks Identifies Enhancer Somatic Mutations That Functionally Impact Lung Cancer.

Enhancers are noncoding regulatory DNA regions that modulate the transcription of target genes, often over large distances along with the genomic sequence. Enhancer alterations have been associated with various pathological conditions, including cancer. However, the identification and characterization of somatic mutations in noncoding regulatory regions with a functional effect on tumorigenesis and prognosis remain a major challenge. Here, we present a strategy for detecting and characterizing enhancer mutations in a genome-wide analysis of patient cohorts, across three lung cancer subtypes. Lung tissue-specific enhancers were defined by integrating experimental data and public epigenomic profiles, and the genome-wide enhancer-target gene regulatory network of lung cells was constructed by integrating chromatin three-dimensional architecture data. Lung cancers possessed a similar mutation burden at tissue-specific enhancers and exons but with differences in their mutation signatures. Functionally relevant alterations were prioritized on the basis of the pathway-level integration of the effect of a mutation and the frequency of mutations on individual enhancers. The genes enriched for mutated enhancers converged on the regulation of key biological processes and pathways relevant to tumor biology. Recurrent mutations in individual enhancers also affected the expression of target genes, with potential relevance for patient prognosis. Together, these findings show that noncoding regulatory mutations have a potential relevance for cancer pathogenesis and can be exploited for patient classification. SIGNIFICANCE: Mapping enhancer-target gene regulatory interactions and analyzing enhancer mutations at the level of their target genes and pathways reveal convergence of recurrent enhancer mutations on biological processes involved in tumorigenesis and prognosis.

La medicina di precisione come opera d'arte collettiva.

Advancing our understanding of complex diseases necessitates an interdisciplinary dialogue beyond artificial intelligence (AI) in the field of medicine. Two decades after the completion of the Human Genome Project, genetic sequencing has facilitated targeted therapies for gene mutation-related ailments. However, this achievement has unveiled the immense gaps in our comprehension of life and disease mechanisms. Complex diseases, including cancer, diabetes, and autoimmune disorders, remain elusive due to their multifactorial nature. Consequently, a more holistic approach integrating AI with diverse scientific disciplines becomes imperative. This paper emphasizes the urgency of fostering collaboration among genetics, molecular biology, computational biology, and clinical research to unravel the intricate complexities underlying these diseases. By synergizing expertise and data from various domains, we can make significant strides towards unraveling the intricate web of complex diseases, leading to improved diagnosis, treatment, and ultimately, patient outcomes.

Network medicine for patients' stratification: From single-layer to multi-omics.

Precision medicine research increasingly relies on the integrated analysis of multiple types of omics. In the era of big data, the large availability of different health-related information represents a great, but at the same time untapped, chance with a potentially fundamental role in the prevention, diagnosis and prognosis of diseases. Computational methods are needed to combine this data to create a comprehensive view of a given disease. Network science can model biomedical data in terms of relationships among molecular players of different nature and has been successfully proposed as a new paradigm for studying human diseases. Patient stratification is an open challenge aimed at identifying subtypes with different disease manifestations, severity, and expected survival time. Several stratification approaches based on high-throughput gene expression measurements have been successfully applied. However, few attempts have been proposed to exploit the integration of various genotypic and phenotypic data to discover novel sub-types or improve the detection of known groupings. This article is categorized under: Cancer > Biomedical Engineering Cancer > Computational Models Cancer > Genetics/Genomics/Epigenetics.

Immune-related toxicity and soluble profile in patients affected by solid tumors: a network approach.

BACKGROUND: Immune checkpoint inhibitors (ICIs) have particular, immune-related adverse events (irAEs), as a consequence of interfering with self-tolerance mechanisms. The incidence of irAEs varies depending on ICI class, administered dose and treatment schedule. The aim of this study was to define a baseline (T0) immune profile (IP) predictive of irAE development. METHODS: A prospective, multicenter study evaluating the immune profile (IP) of 79 patients with advanced cancer and treated with anti-programmed cell death protein 1 (anti-PD-1) drugs as a first- or second-line setting was performed. The results were then correlated with irAEs onset. The IP was studied by means of multiplex assay, evaluating circulating concentration of 12 cytokines, 5 chemokines, 13 soluble immune checkpoints and 3 adhesion molecules. Indoleamine 2, 3-dioxygenase (IDO) activity was measured through a modified liquid chromatography-tandem mass spectrometry using the high-performance liquid chromatography-mass spectrometry (HPLC-MS/MS) method. A connectivity heatmap was obtained by calculating Spearman correlation coefficients. Two different networks of connectivity were constructed, based on the toxicity profile. RESULTS: Toxicity was predominantly of low/moderate grade. High-grade irAEs were relatively rare, while cumulative toxicity was high (35%). Positive and statistically significant correlations between the cumulative toxicity and IP10 and IL8, sLAG3, sPD-L2, sHVEM, sCD137, sCD27 and sICAM-1 serum concentration were found. Moreover, patients who experienced irAEs had a markedly different connectivity pattern, characterized by disruption of most of the paired connections between cytokines, chemokines and connections of sCD137, sCD27 and sCD28, while sPDL-2 pair-wise connectivity values seemed to be intensified. Network connectivity analysis identified a total of 187 statistically significant interactions in patients without toxicity and a total of 126 statistically significant interactions in patients with toxicity. Ninety-eight interactions were common to both networks, while 29 were specifically observed in patients who experienced toxicity. CONCLUSIONS: A particular, common pattern of immune dysregulation was defined in patients developing irAEs. This immune serological profile, if confirmed in a larger patient population, could lead to the design of a personalized therapeutic strategy in order to prevent, monitor and treat irAEs at an early stage.

A Comparison of Network-Based Methods for Drug Repurposing along with an Application to Human Complex Diseases.

Drug repurposing strategy, proposing a therapeutic switching of already approved drugs with known medical indications to new therapeutic purposes, has been considered as an efficient approach to unveil novel drug candidates with new pharmacological activities, significantly reducing the cost and shortening the time of de novo drug discovery. Meaningful computational approaches for drug repurposing exploit the principles of the emerging field of Network Medicine, according to which human diseases can be interpreted as local perturbations of the human interactome network, where the molecular determinants of each disease (disease genes) are not randomly scattered, but co-localized in highly interconnected subnetworks (disease modules), whose perturbation is linked to the pathophenotype manifestation. By interpreting drug effects as local perturbations of the interactome, for a drug to be on-target effective against a specific disease or to cause off-target adverse effects, its targets should be in the nearby of disease-associated genes. Here, we used the network-based proximity measure to compute the distance between the drug module and the disease module in the human interactome by exploiting five different metrics (minimum, maximum, mean, median, mode), with the aim to compare different frameworks for highlighting putative repurposable drugs to treat complex human diseases, including malignant breast and prostate neoplasms, schizophrenia, and liver cirrhosis. Whilst the standard metric (that is the minimum) for the network-based proximity remained a valid tool for efficiently screening off-label drugs, we observed that the other implemented metrics specifically predicted further interesting drug candidates worthy of investigation for yielding a potentially significant clinical benefit.

Convolutional Neural Networks for Automated Classification of Prostate Multiparametric Magnetic Resonance Imaging Based on Image Quality.

BACKGROUND: Prostate magnetic resonance imaging (MRI) is technically demanding, requiring high image quality to reach its full diagnostic potential. An automated method to identify diagnostically inadequate images could help optimize image quality. PURPOSE: To develop a convolutional neural networks (CNNs) based analysis pipeline for the classification of prostate MRI image quality. STUDY TYPE: Retrospective. SUBJECTS: Three hundred sixteen prostate mpMRI scans and 312 men (median age 67). FIELD STRENGTH/SEQUENCE: A 3 T; fast spin echo T2WI, echo planar imaging DWI, ADC, gradient-echo dynamic contrast enhanced (DCE). ASSESSMENT: MRI scans were reviewed by three genitourinary radiologists (V.P., M.D.M., S.C.) with 21, 12, and 5 years of experience, respectively. Sequences were labeled as high quality (Q1) or low quality (Q0) and used as the reference standard for all analyses. STATISTICAL TESTS: Sequences were split into training, validation, and testing sets (869, 250, and 120 sequences, respectively). Inter-reader agreement was assessed with the Fleiss kappa. Following preprocessing and data augmentation, 28 CNNs were trained on MRI slices for each sequence. Model performance was assessed on both a per-slice and a per-sequence basis. A pairwise t-test was performed to compare performances of the classifiers. RESULTS: The number of sequences labeled as Q0 or Q1 was 38 vs. 278 for T2WI, 43 vs. 273 for DWI, 41 vs. 275 for ADC, and 38 vs. 253 for DCE. Inter-reader agreement was almost perfect for T2WI and DCE and substantial for DWI and ADC. On the per-slice analysis, accuracy was 89.95% ± 0.02% for T2WI, 79.83% ± 0.04% for DWI, 76.64% ± 0.04% for ADC, 96.62% ± 0.01% for DCE. On the per-sequence analysis, accuracy was 100% ± 0.00% for T2WI, DWI, and DCE, and 92.31% ± 0.00% for ADC. The three best algorithms performed significantly better than the remaining ones on every sequence (P-value < 0.05). DATA CONCLUSION: CNNs achieved high accuracy in classifying prostate MRI image quality on an individual-slice basis and almost perfect accuracy when classifying the entire sequences. EVIDENCE LEVEL: 4 TECHNICAL EFFICACY: Stage 1.

Network Analysis Integrating microRNA Expression Profiling with MRI Biomarkers and Clinical Data for Prostate Cancer Early Detection: A Proof of Concept Study.

The MRI of the prostate is the gold standard for the detection of clinically significant prostate cancer (csPCa). Nonetheless, MRI still misses around 11% of clinically significant disease. The aim was to comprehensively integrate tissue and circulating microRNA profiling, MRI biomarkers and clinical data to implement PCa early detection. In this prospective cohort study, 76 biopsy naïve patients underwent MRI and MRI directed biopsy. A sentinel sample of 15 patients was selected for a pilot molecular analysis. Weighted gene coexpression network analysis was applied to identify the microRNAs drivers of csPCa. MicroRNA-target gene interaction maps were constructed, and enrichment analysis performed. The ANOVA on ranks test and ROC analysis were performed for statistics. Disease status was associated with the underexpression of the miRNA profiled; a correlation was found with ADC (r = -0.51, p = 0.02) and normalized ADC values (r = -0.64, p = 0.002). The overexpression of miRNAs from plasma was associated with csPCa (r = 0.72; p = 0.02), and with PI-RADS assessment score (r = 0.73; p = 0.02); a linear correlation was found with biomarkers of diffusion and perfusion. Among the 800 profiled microRNA, eleven were identified as correlating with PCa, among which miR-548a-3p, miR-138-5p and miR-520d-3p were confirmed using the RT-qPCR approach on an additional cohort of ten subjects. ROC analysis showed an accuracy of >90%. Provided an additional validation set of the identified miRNAs on a larger cohort, we propose a diagnostic paradigm shift that sees molecular data and MRI biomarkers as the prebiopsy triage of patients at risk for PCa. This approach will allow for accurate patient allocation to biopsy, and for stratification into risk group categories, reducing overdiagnosis and overtreatment.

The role of FOSL1 in stem-like cell reprogramming processes.

Cancer stem-like cells (CSCs) have self-renewal abilities responsible for cancer progression, therapy resistance, and metastatic growth. The glioblastoma stem-like cells are the most studied among CSC populations. A recent study identified four transcription factors (SOX2, SALL2, OLIG2, and POU3F2) as the minimal core sufficient to reprogram differentiated glioblastoma (GBM) cells into stem-like cells. Transcriptomic data of GBM tissues and cell lines from two different datasets were then analyzed by the SWItch Miner (SWIM), a network-based software, and FOSL1 was identified as a putative regulator of the previously identified minimal core. Herein, we selected NTERA-2 and HEK293T cells to perform an in vitro study to investigate the role of FOSL1 in the reprogramming mechanisms. We transfected the two cell lines with a constitutive FOSL1 cDNA plasmid. We demonstrated that FOSL1 directly regulates the four transcription factors binding their promoter regions, is involved in the deregulation of several stemness markers, and reduces the cells' ability to generate aggregates increasing the extracellular matrix component FN1. Although further experiments are necessary, our data suggest that FOSL1 reprograms the stemness by regulating the core of the four transcription factors.

SEaCorAl: Identifying and contrasting the regulation-correlation bias in RNA-Seq paired expression data of patient groups.

The Cancer Genome Atlas database offers the possibility of analyzing genome-wide expression RNA-Seq cancer data using paired counts, that is, studies where expression data are collected in pairs of normal and cancer cells, by taking samples from the same individual. Correlation of gene expression profiles is the most common analysis to study co-expression groups, which is used to find biological interpretation of -omics big data. The aim of the paper is threefold: firstly we show for the first time, the presence of a "regulation-correlation bias" in RNA-Seq paired expression data, that is an artifactual link between the expression status (up- or down-regulation) of a gene pair and the sign of the corresponding correlation coefficient. Secondly, we provide a statistical model able to theoretically explain the reasons for the presence of such a bias. Thirdly, we present a bias-removal algorithm, called SEaCorAl, able to effectively reduce bias effects and improve the biological significance of correlation analysis. Validation of the SEaCorAl algorithm is performed by showing a significant increase in the ability to detect biologically meaningful associations of positive correlations and a significant increase of the modularity of the resulting unbiased correlation network.

Prostate cancer screening research can benefit from network medicine: an emerging awareness.

Up to date, screening for prostate cancer (PCa) remains one of the most appealing but also a very controversial topics in the urological community. PCa is the second most common cancer in men worldwide and it is universally acknowledged as a complex disease, with a multi-factorial etiology. The pathway of PCa diagnosis has changed dramatically in the last few years, with the multiparametric magnetic resonance (mpMRI) playing a starring role with the introduction of the "MRI Pathway". In this scenario the basic tenet of network medicine (NM) that sees the disease as perturbation of a network of interconnected molecules and pathways, seems to fit perfectly with the challenges that PCa early detection must face to advance towards a more reliable technique. Integration of tests on body fluids, tissue samples, grading/staging classification, physiological parameters, MR multiparametric imaging and molecular profiling technologies must be integrated in a broader vision of "disease" and its complexity with a focus on early signs. PCa screening research can greatly benefit from NM vision since it provides a sound interpretation of data and a common language, facilitating exchange of ideas between clinicians and data analysts for exploring new research pathways in a rational, highly reliable, and reproducible way.

Connectivity Significance for Disease Gene Prioritization in an Expanding Universe.

A fundamental topic in network medicine is disease genes prioritization. The underlying hypothesis is that disease genes are organized as modules confined within the interactome. Here, we propose a novel algorithm called DiaBLE (DIAMOnD Background Local Expansion) which is a modified version of DIAMOnD, a successful algorithm based on the concept of connectivity significance. Instead of taking the whole interactome as the background model, DiaBLE considers as gene universe the smallest local expansion of the current seeds set at each iteration step. We show that DiaBLE significantly increases the overall DIAMOnD ranking quality of genes prioritization both in terms of cross-validation and biological consistency. Here, we focus on the two algorithms only since a comparative analysis among gene prioritization methods is beyond the scope of this study. Finally, we briefly discuss the improvement of biological insight provided by DiaBLE for two cancers (head and neck squamous cell carcinoma and kidney renal clear cell carcinoma).

Identification of Disease-miRNA Networks Across Different Cancer Types Using SWIM.

MicroRNAs (miRNAs) are small noncoding RNAs (ncRNAs) involved in several biological processes and diseases. MiRNAs regulate gene expression at the posttranscriptional level, mostly downregulating their targets by binding specific regions of transcripts through imperfect sequence complementarity. Prediction of miRNA-binding sites is challenging, and target prediction algorithms are usually based on sequence complementarity. In the last years, it has been shown that by adding miRNA and protein coding gene expression, we are able to build tissue-, cell line-, or disease-specific networks improving our understanding of complex biological scenarios. In this chapter, we present an application of a recently published software named SWIM, that allows to identify key genes in a network of interactions by defining appropriate "roles" of genes according to their local/global positioning in the overall network. Furthermore, we show how the SWIM software can be used to build miRNA-disease networks, by applying the approach to tumor data obtained from The Cancer Genome Atlas (TCGA).

BRAFV600E-mutant cancers display a variety of networks by SWIM analysis: prediction of vemurafenib clinical response.

PURPOSE: Several studies have shown that different tumour types sharing a driver gene mutation do not respond uniformly to the same targeted agent. Our aim was to use an unbiased network-based approach to investigate this fundamental issue using BRAFV600E mutant tumours and the BRAF inhibitor vemurafenib. METHODS: We applied SWIM, a software able to identify putative regulatory (switch) genes involved in drastic changes to the cell phenotype, to gene expression profiles of different BRAFV600E mutant cancers and their normal counterparts in order to identify the switch genes that could potentially explain the heterogeneity of these tumours' responses to vemurafenib. RESULTS: We identified lung adenocarcinoma as the tumour with the highest number of switch genes (298) compared to its normal counterpart. By looking for switch genes encoding for kinases with homology sequences similar to known vemurafenib targets, we found that thyroid cancer and lung adenocarcinoma have a similar number of putative targetable switch gene kinases (5 and 6, respectively) whereas colorectal cancer has just one. CONCLUSIONS: We are persuaded that our network analysis may aid in the comprehension of molecular mechanisms underlying the different responses to vemurafenib in BRAFV600E mutant tumours.

A feature-based integrated scoring scheme for cell cycle-regulated genes prioritization.

Prioritization of cell cycle-regulated genes from expression time-profiles is still an open problem. The point at issue is the surprisingly poor overlap among ranked lists obtained from different experimental protocols. Instead of developing a general-purpose computational methodology for detecting periodic signals, we focus on the budding yeast mitotic cell cycle. The reason being that the current availability of a total of 12 datasets, produced by 6 independent groups using 4 different synchronization methods, permits a re-analysis and re-consideration of this problem in a more reliable and extensive data domain. Notably, budding yeast is a model organism for studying cancer and testing new drugs. Here we propose a novel multi-feature score (called PERLA, PERiodicity, Regulation and Lag-Autocorrelation) that integrates different features of cell cycle-regulated gene expression time-profiles. We obtained increased performances on a wide range of benchmarks and, most importantly, a substantially increased overlap of the top ranking genes among different datasets, thus proving the effectiveness of the proposed prioritization algorithm. Examples on how to use PERLA to gain new insight into the biology of the cell cycle, are provided in a final dedicated section.

Network-Based Approaches to Explore Complex Biological Systems towards Network Medicine.

Network medicine relies on different types of networks: from the molecular level of protein⁻protein interactions to gene regulatory network and correlation studies of gene expression. Among network approaches based on the analysis of the topological properties of protein⁻protein interaction (PPI) networks, we discuss the widespread DIAMOnD (disease module detection) algorithm. Starting from the assumption that PPI networks can be viewed as maps where diseases can be identified with localized perturbation within a specific neighborhood (i.e., disease modules), DIAMOnD performs a systematic analysis of the human PPI network to uncover new disease-associated genes by exploiting the connectivity significance instead of connection density. The past few years have witnessed the increasing interest in understanding the molecular mechanism of post-transcriptional regulation with a special emphasis on non-coding RNAs since they are emerging as key regulators of many cellular processes in both physiological and pathological states. Recent findings show that coding genes are not the only targets that microRNAs interact with. In fact, there is a pool of different RNAs-including long non-coding RNAs (lncRNAs) -competing with each other to attract microRNAs for interactions, thus acting as competing endogenous RNAs (ceRNAs). The framework of regulatory networks provides a powerful tool to gather new insights into ceRNA regulatory mechanisms. Here, we describe a data-driven model recently developed to explore the lncRNA-associated ceRNA activity in breast invasive carcinoma. On the other hand, a very promising example of the co-expression network is the one implemented by the software SWIM (switch miner), which combines topological properties of correlation networks with gene expression data in order to identify a small pool of genes-called switch genes-critically associated with drastic changes in cell phenotype. Here, we describe SWIM tool along with its applications to cancer research and compare its predictions with DIAMOnD disease genes.

SWIM: a computational tool to unveiling crucial nodes in complex biological networks.

SWItchMiner (SWIM) is a wizard-like software implementation of a procedure, previously described, able to extract information contained in complex networks. Specifically, SWIM allows unearthing the existence of a new class of hubs, called "fight-club hubs", characterized by a marked negative correlation with their first nearest neighbors. Among them, a special subset of genes, called "switch genes", appears to be characterized by an unusual pattern of intra- and inter-module connections that confers them a crucial topological role, interestingly mirrored by the evidence of their clinic-biological relevance. Here, we applied SWIM to a large panel of cancer datasets from The Cancer Genome Atlas, in order to highlight switch genes that could be critically associated with the drastic changes in the physiological state of cells or tissues induced by the cancer development. We discovered that switch genes are found in all cancers we studied and they encompass protein coding genes and non-coding RNAs, recovering many known key cancer players but also many new potential biomarkers not yet characterized in cancer context. Furthermore, SWIM is amenable to detect switch genes in different organisms and cell conditions, with the potential to uncover important players in biologically relevant scenarios, including but not limited to human cancer.

Role of the long non-coding RNA PVT1 in the dysregulation of the ceRNA-ceRNA network in human breast cancer.

Recent findings have identified competing endogenous RNAs (ceRNAs) as the drivers in many disease conditions, including cancers. The ceRNAs indirectly regulate each other by reducing the amount of microRNAs (miRNAs) available to target messenger RNAs (mRNAs). The ceRNA interactions mediated by miRNAs are modulated by a titration mechanism, i.e. large changes in the ceRNA expression levels either overcome, or relieve, the miRNA repression on competing RNAs; similarly, a very large miRNA overexpression may abolish competition. The ceRNAs are also called miRNA "decoys" or miRNA "sponges" and encompass different RNAs competing with each other to attract miRNAs for interactions: mRNA, long non-coding RNAs (lncRNAs), pseudogenes, or circular RNAs. Recently, we developed a computational method for identifying ceRNA-ceRNA interactions in breast invasive carcinoma. We were interested in unveiling which lncRNAs could exert the ceRNA activity. We found a drastic rewiring in the cross-talks between ceRNAs from the physiological to the pathological condition. The main actor of this dysregulated lncRNA-associated ceRNA network was the lncRNA PVT1, which revealed a net biding preference towards the miR-200 family members in normal breast tissues. Despite its up-regulation in breast cancer tissues, mimicked by the miR-200 family members, PVT1 stops working as ceRNA in the cancerous state. The specific conditions required for a ceRNA landscape to occur are still far from being determined. Here, we emphasized the importance of the relative concentration of the ceRNAs, and their related miRNAs. In particular, we focused on the withdrawal in breast cancer tissues of the PVT1 ceRNA activity and performed a gene expression and sequence analysis of its multiple isoforms. We found that the PVT1 isoform harbouring the binding site for a representative miRNA of the miR-200 family shows a drastic decrease in its relative concentration with respect to the miRNA abundance in breast cancer tissues, providing a plausibility argument to the breakdown of the sponge program orchestrated by the oncogene PVT1.

PVT1: a rising star among oncogenic long noncoding RNAs.

It is becoming increasingly clear that short and long noncoding RNAs critically participate in the regulation of cell growth, differentiation, and (mis)function. However, while the functional characterization of short non-coding RNAs has been reaching maturity, there is still a paucity of well characterized long noncoding RNAs, even though large studies in recent years are rapidly increasing the number of annotated ones. The long noncoding RNA PVT1 is encoded by a gene that has been long known since it resides in the well-known cancer risk region 8q24. However, a couple of accidental concurrent conditions have slowed down the study of this gene, that is, a preconception on the primacy of the protein-coding over noncoding RNAs and the prevalent interest in its neighbor MYC oncogene. Recent studies have brought PVT1 under the spotlight suggesting interesting models of functioning, such as competing endogenous RNA activity and regulation of protein stability of important oncogenes, primarily of the MYC oncogene. Despite some advancements in modelling the PVT1 role in cancer, there are many questions that remain unanswered concerning the precise molecular mechanisms underlying its functioning.

Computational analysis identifies a sponge interaction network between long non-coding RNAs and messenger RNAs in human breast cancer.

BACKGROUND: Non-coding RNAs (ncRNAs) are emerging as key regulators of many cellular processes in both physiological and pathological states. Moreover, the constant discovery of new non-coding RNA species suggests that the study of their complex functions is still in its very early stages. This variegated class of RNA species encompasses the well-known microRNAs (miRNAs) and the most recently acknowledged long non-coding RNAs (lncRNAs). Interestingly, in the last couple of years, a few studies have shown that some lncRNAs can act as miRNA sponges, i.e. as competing endogenous RNAs (ceRNAs), able to reduce the amount of miRNAs available to target messenger RNAs (mRNAs). RESULTS: We propose a computational approach to explore the ability of lncRNAs to act as ceRNAs by protecting mRNAs from miRNA repression. A seed match analysis was performed to validate the underlying regression model. We built normal and cancer networks of miRNA-mediated sponge interactions (MMI-networks) using breast cancer expression data provided by The Cancer Genome Atlas. CONCLUSIONS: Our study highlights a marked rewiring in the ceRNA program between normal and pathological breast tissue, documented by its "on/off" switch from normal to cancer, and vice-versa. This mutually exclusive activation confers an interesting character to ceRNAs as potential oncosuppressive, or oncogenic, protagonists in cancer. At the heart of this phenomenon is the lncRNA PVT1, as illustrated by both the width of its antagonist mRNAs in normal-MMI-network, and the relevance of the latter in breast cancer. Interestingly, PVT1 revealed a net binding preference towards the mir-200 family as the bone of contention with its rival mRNAs.