Search

Motivation The principle of Breiman's random forest (RF) is to build and assemble complementary classification trees in a way that maximizes their variability. We propose a new type of random forest that disobeys Breiman's principles and involves building trees with no classification errors in very large quantities. We used a new type of decision tree that uses a neuron at each node as well as an in-innovative half Christmas tree structure. With these new RFs, we developed a score, based on a family of ten new statistical information criteria, called Nguyen information criteria (NICs), to evaluate the predictive qualities of features in three dimensions. Results The first NIC allowed the Akaike information criterion to be minimized more quickly than data obtained with the Gini index when the features were introduced in a logistic regression model. The selected features based on the NICScore showed a slight advantage compared to the support vector machines—recursive feature elimination (SVM-RFE) method. We demonstrate that the inclusion of artificial neurons in tree nodes allows a large number of classifiers in the same node to be taken into account simultaneously and results in perfect trees without classification errors. Availability and implementation The methods used to build the perfect trees in this article were implemented in the 'ROP' R package, archived at https://cran.r-project.org/web/packages/ROP/index.html. Supplementary information Supplementary data are available at Bioinformatics online. [ABSTRACT FROM AUTHOR]

This paper is devoted to the estimation of a partial graphical model with a structural Bayesian penalization. Precisely, we are interested in the linear regression setting where the estimation is made through the direct links between potentially high-dimensional predictors and multiple responses, since it is known that Gaussian graphical models enable to exhibit direct links only, whereas coefficients in linear regressions contain both direct and indirect relations (due e.g. to strong correlations among the variables). A smooth penalty reflecting a generalized Gaussian Bayesian prior on the covariates is added, either enforcing patterns (like row structures) in the direct links or regulating the joint influence of predictors. We give a theoretical guarantee for our method, taking the form of an upper bound on the estimation error arising with high probability, provided that the model is suitably regularized. Empirical studies on synthetic data and a real dataset are conducted. [ABSTRACT FROM AUTHOR]

Background: Breast cancer is the most common cancer in women and the first cancer concerning mortality. Metastatic breast cancer remains a disease with a poor prognosis and about 30% of women diagnosed with an early stage will have a secondary progression. Metastatic breast cancer is an incurable disease despite significant therapeutic advances in both supportive cares and targeted specific therapies. In the management of a metastatic patient, each clinician follows a highly complex and strictly personal decision making process. It is based on a number of objective and subjective parameters which guides therapeutic choice in the most individualized or adapted manner.Methods/design: The main objective is to integrate massive and heterogeneous data concerning the patient's environment, personal and familial history, clinical and biological data, imaging, histological results (with multi-omics data), and microbiota analysis. These characteristics are multiple and in dynamic interaction overtime. With the help of mathematical units with biological competences and scientific collaborations, our project is to improve the comprehension of treatment response, based on health clinical and molecular heterogeneous big data investigation.Discussion: Our project is to prove feasibility of creation of a clinico-biological database prospectively by collecting epidemiological, socio-economic, clinical, biological, pathological, multi-omic data and to identify characteristics related to the overall survival status before treatment and within 15 years after treatment start from a cohort of 300 patients with a metastatic breast cancer treated in the institution.Trial Registration: ClinicalTrials.gov identifier (NCT number): NCT03958136 . Registration 21st of May, 2019; retrospectively registered. [ABSTRACT FROM AUTHOR]

'Breast cancer gene-expression miner' (bc-GenExMiner) is a breast cancer–associated web portal (http://bcgenex.ico.unicancer.fr). Here, we describe the development of a new statistical mining module, which permits several differential gene expression analyses, i.e. 'Expression' module. Sixty-two breast cancer cohorts and one healthy breast cohort with their corresponding clinicopathological information are included in bc-GenExMiner v4.5 version. Analyses are based on microarray or RNAseq transcriptomic data. Thirty-nine differential gene expression analyses, grouped into 13 categories, according to clinicopathological and molecular characteristics ('Targeted' and 'Exhaustive') and gene expression ('Customized'), have been developed. Output results are visualized in four forms of plots. This new statistical mining module offers, among other things, the possibility to compare gene expression in healthy (cancer-free), tumour-adjacent and tumour tissues at once and in three triple-negative breast cancer subtypes (i.e. C1: molecular apocrine tumours; C2: basal-like tumours infiltrated by immune suppressive cells and C3: basal-like tumours triggering an ineffective immune response). Several validation tests showed that bioinformatics process did not alter the pathobiological information contained in the source data. In this work, we developed and demonstrated that bc-GenExMiner 'Expression' module can be used for exploratory and validation purposes. Database URL : http://bcgenex.ico.unicancer.fr [ABSTRACT FROM AUTHOR]

This paper is devoted to the estimation of a partial graphical model with a structural Bayesian penalization. Precisely, we are interested in the linear regression setting where the estimation is made through the direct links between potentially high-dimensional predictors and multiple responses, since it is known that Gaussian graphical models enable to exhibit direct links only, whereas coefficients in linear regressions contain both direct and indirect relations (due e.g. to strong correlations among the variables). A smooth penalty reflecting a generalized Gaussian Bayesian prior on the covariates is added, either enforcing patterns (like row structures) in the direct links or regulating the joint influence of predictors. We give a theoretical guarantee for our method, taking the form of an upper bound on the estimation error arising with high probability, provided that the model is suitably regularized. Empirical studies on synthetic data and a real dataset are conducted. [ABSTRACT FROM AUTHOR]

Background: Breast cancer biological characteristics change as age advances. Today, there is a lack of knowledge regarding age-specific molecular alterations that characterize breast tumours, notably in elderly patients. The vast majority of studies that aimed at exploring breast cancer in function of age are based on clinico-pathological data. Gene-expression signatures (GES), which in some ways capture biological information in a non-reductionist manner, represent powerful tools able to explore tumour heterogeneity. Methods: Twenty-five GES were used for functional annotation of breast tumours in function of age: five for molecular subtyping, seven for immune response, three for metabolism, seven for critical pathways in cancer and three for prognosis. Affymetrix® genomics datasets were exclusively used to avoid cross-platform normalization issues. Available corresponding clinico-pathological data were also retrieved and analysed. Results: Fifteen publicly available datasets were pooled for a total of 2378 breast cancer patients (whole cohort), out of whom 1413 were of Caucasian origin. Three age groups were defined: ≤ 40 years (AG1), > 40 to < 70 years (AG2) and ≥ 70 years (AG3). We confirmed that age influenced the incidence of molecular subtypes. We found a significant growing incidence of luminal B and a decreasing kinetics for basal-like in function of age. We showed that AG3 luminal B tumours were less aggressive than AG1 luminal B tumours based on different GES (iron metabolism, mitochondrial oxidative phosphorylation and reactive stroma), recurrence score prognostic GES and histological grade (SBR). Contrary to tumours of young patients, tumours of elderly patients concentrated favourable GES scores: high oestrogen receptor and mitochondrial oxidative phosphorylation, low proliferation, basal-like, glycolysis, chromosomal instability and iron metabolism, and low GES prognostic scores (van't Veer 70-GES, genomic grade index and recurrence score). Conclusions: Functional annotation of breast tumours by means of 25 GES demonstrated a decreasing aggressiveness of breast tumours in function of age. This strategy, which can be strengthened by increasing the number of representative GES to gain more insight into biological systems involved in this disease, provides a framework to develop rational therapeutic strategies in function of age. [ABSTRACT FROM AUTHOR]

We recently developed a user-friendly web-based application called bc-GenExMiner (http://bcgenex.centregauducheau.fr), which offered the possibility to evaluate prognostic informativity of genes in breast cancer by means of a 'prognostic module'. In this study, we develop a new module called 'correlation module', which includes three kinds of gene expression correlation analyses. The first one computes correlation coefficient between 2 or more (up to 10) chosen genes. The second one produces two lists of genes that are most correlated (positively and negatively) to a 'tested' gene. A gene ontology (GO) mining function is also proposed to explore GO 'biological process', 'molecular function' and 'cellular component' terms enrichment for the output lists of most correlated genes. The third one explores gene expression correlation between the 15 telomeric and 15 centromeric genes surrounding a 'tested' gene. These correlation analyses can be performed in different groups of patients: all patients (without any subtyping), in molecular subtypes (basal-like, HER2+, luminal A and luminal B) and according to oestrogen receptor status. Validation tests based on published data showed that these automatized analyses lead to results consistent with studies' conclusions. In brief, this new module has been developed to help basic researchers explore molecular mechanisms of breast cancer. [ABSTRACT FROM AUTHOR]

Background: Copy number alterations (CNA) are acquired during the evolution of cancers from their early stage to metastatic stage. This study aims at analysing the clinical value of the identified metastasis-associated CNAs both in metastatic breast cancers (mBCs) and early breast cancers (eBCs). Methods: Single-nucleotide polymorphism (SNP)-array was performed on 926 biopsies from mBC patients, enrolled in SAFIR02-BREAST prospective trial. CNA profiles of eBCs from The Cancer Genome Atlas Breast Invasive Carcinoma (n = 770), Molecular Taxonomy of Breast Cancer International Consortium (n = 1620) and PACS04 trial (n = 243) cohorts were used as references for comparing mBCs and eBCs CNA profiles. Overall survival was the considered survival endpoint. Results: Among the twenty-one genes frequently altered in ER + /HER2− mBCs: focal amplification of TERT was associated with poor outcome in the ER + /HER2− mBC population. Among the ER + /HER2− mBCs patients for whom CDK4/6 inhibitors information before biopsies collection was available: we identified seven genes on post-treatment biopsies, including the cyclin-dependent kinase 4 (CDK4), which was amplified in 9.8% of the ER + /HER2− mBCs pretreated population, as compared to 1.5% in the ER + /HER2− mBCs unpretreated population (P = 2.82E-04) as well as the 3 eBC populations. CDK4 amplification was associated with poor outcome in the ER + /HER2− eBCs. Conclusions: This study provides insights into the biology of mBCs and identifies clinically useful genomic features for future improvement of breast cancer patient management. [ABSTRACT FROM AUTHOR]

Proteomics has been little used for the identification of novel prognostic and/or therapeutic markers in isocitrate dehydrogenase (IDH)‐wildtype glioblastoma (GB). In this study, we analyzed 50 tumor and 30 serum samples from short‐ and long‐term survivors of IDH‐wildtype GB (STS and LTS, respectively) by data‐independent acquisition mass spectrometry (DIA‐MS)‐based proteomics, with the aim of identifying such markers. DIA‐MS identified 5422 and 826 normalized proteins in tumor and serum samples, respectively, with only three tumor proteins and 26 serum proteins displaying significant differential expression between the STS and LTS groups. These dysregulated proteins were principally associated with the detoxification of reactive oxygen species (ROS). In particular, GB patients in the STS group had high serum levels of malate dehydrogenase 1 (MDH1) and ribonuclease inhibitor 1 (RNH1) and low tumor levels of fatty acid‐binding protein 7 (FABP7), which may have enabled them to maintain low ROS levels, counteracting the effects of the first‐line treatment with radiotherapy plus concomitant and adjuvant temozolomide. A blood score built on the levels of MDH1 and RNH1 expression was found to be an independent prognostic factor for survival based on the serum proteome data for a cohort of 96 IDH‐wildtype GB patients. This study highlights the utility of circulating MDH1 and RNH1 biomarkers for determining the prognosis of patients with IDH‐wildtype GB. Furthermore, the pathways driven by these biomarkers, and the tumor FABP7 pathway, may constitute promising therapeutic targets for blocking ROS detoxification to overcome resistance to chemoradiotherapy in potential GB STS. [ABSTRACT FROM AUTHOR]

Simple Summary: HER2-low (IHC 1+ or IHC 2+/ISH-) breast cancer (BC) is an emerging subtype of BC with promising results with antibody drug conjugate (ADC) in the metastatic setting. In the early setting, few data have been reported regarding the predictive and prognostic impact of HER2-low status in triple-negative BC (TNBC). Purpose: Investigates the link between HER2 status and histological response after neoadjuvant chemotherapy in patients with early TNBC. Methods: We retrieved clinical and anatomopathological data retrospectively from 449 patients treated for the first time with standard neoadjuvant chemotherapy for early unilateral BC between 2005 and 2020. The primary endpoint was pathological complete response (pCR, i.e., ypT0 ypN0), according to HER2 status. Secondary endpoints included invasive disease-free survival (I-DFS) and overall survival (OS). Results: 437 patients were included, and 121 (27.7%) patients had HER2-low tumours. The pCR rate was not significantly different between the HER2-low group vs. the HER2-0 group (35.7% versus 41.8%, p = 0.284) in either univariate analysis or multivariate analysis adjusted for TNM classification and grade (odds ratio [OR] = 0.70, confidence interval [CI] 95% 0.45–1.08). With a median follow-up of 72.9 months, no significant survival differences were observed between patients with HER2-low tumours vs. patients with HER2-0 tumours in terms of I-DFS (p = 0.487) and OS (p = 0.329). Conclusions: In our cohort, HER2 status was not significantly associated with pCR in a manner consistent with data published recently on TNBC. However, the prognostic impact of HER2-low expression among TNBC patients warrants further evaluation. [ABSTRACT FROM AUTHOR]

Breast cancer may present genomic alterations leading to homologous recombination deficiency (HRD). PARP inhibitors have proven their efficacy in patients with HER2-negative (HER2-) metastatic breast cancer (mBC) harbouring germline (g) BRCA1/2 mutations in 3 phases III trials. The single-arm phase II RUBY trial included 42 patients, 40 of whom received at least one dose of rucaparib. RUBY study assessed the efficacy of rucaparib in HER2-mBC with either high genomic loss of heterozygosity (LOH) score or non-germline BRCA1/2 mutation. The primary objective was the clinical benefit rate (CBR), and the study was powered to see 20% CBR using a 2-stage Simon design. The primary-end point was not reached with a CBR of 13.5%. Two LOH-high patients, without somatic BRCA1/2 mutation, presented a complete and durable response (12 and 28.5 months). Whole-genome analysis was performed on 24 samples, including 5 patients who presented a clinical benefit from rucaparib. HRDetect tended to be associated with response to rucaparib, without reaching statistical significance (median HRDetect responders versus non-responders: 0.465 versus 0.040; p = 0.2135). Finally, 220 of 711 patients with mBC screened for LOH upstream from RUBY presented a high LOH score associated with a higher likelihood of death (hazard ratio = 1.39; 95% CI: 1.11–1.75; p = 0.005). Our data suggest that a small subset of patients with high LOH scores without germline BRCA1/2 mutation could derive benefit from PARP inhibitors. However, the RUBY study underlines the need to develop additional biomarkers to identify selectively potential responders. • Homologous recombination deficiency (HRD) profile can be assessed by genomic scars as genome-wide loss of heterozygosity or HRDetect. • Four patients harbouring high genomic loss of heterozygosity score benefited from rucaparib. • Two with either gPALB2 and/or sBRCA1/2 presented at least a stable disease under rucaparib. • The predictive value of HRDetect to rucaparib was explored with efficiency signals. [ABSTRACT FROM AUTHOR]

Background: The standard therapy regimen of conventional osteosarcoma includes neoadjuvant chemotherapy followed by surgical resection and postoperative chemotherapy. The percentage of necrotic tissue following induction chemotherapy is assessed by using the Huvos grading system, which classifies patients as "poor responders" (PR) and "good responders" (GR). The aim of this study was to identify molecular markers expressed differentially between good and poor responders to neoadjuvant chemotherapy in order to predict the response to chemotherapy in conventional osteosarcomas before beginning treatment. Methods: Suppression Substractive Hybridization (SSH) was performed by using cDNA from frozen biopsy specimens. Expression of selected relevant genes identified by SSH was validated by using QRT-PCR. Immunohistochemistry (IHC) on tissue microarray (TMA) sections of 52 biopsies was performed to investigate protein expression in an independent cohort. Results: ERK1 and STAT3 mRNA level were significantly different between PR and GR in an independent cohort. Phosphorylated STAT3 and ERK1 expressions by IHC on TMA were correlated with poor response to chemotherapy. Conclusions: Our results suggest that ERK1 and STAT3 expression are good predictive markers for chemotherapy response and that inhibitors might be used in combination with common chemotherapeutic drugs in conventional osteosarcomas. [ABSTRACT FROM AUTHOR]

Induction of epithelial-to-mesenchymal transition (EMT) by TGF-β1 requires Ras signaling. We recently identified the transcriptional repressor Blimp-1 (PRDM1) as a downstream effector of the NF-βB, RelB/Bcl-2/Rasdriven pathway that promotes breast cancer cell migration. As the RelB/Blimp-1 pathway similarly required Ras signaling activation, we tested whether Blimp-1 plays a role in TGF-β1--mediated EMT. Here, TGF-β1 treatment of untransformed NMuMG mammary epithelial and MDA-MB-231 breast cancer cells was shown to induce Blimp-1 expression, which promoted an EMT signature and cell migration. TGFB1 and BLIMP1 RNA levels were correlated in patient breast tumors. BLIMP1 gene transcription was activated by TGF-β1 via a c-Raf (RAF1) to AP-1 pathway. Blimp-1 induced expression of the EMT master regulator Snail (SNAI1) via repressing BMP-5, which inhibited Snail expression upon TGF-β1 treatment. Interestingly, a similar cascade was observed during postnatal mouse mammary gland development. RelB expression was detected early in pregnancy followed progressively by Blimp-1 and then Snail; whereas, BMP-5 levels were high in nulliparous and regressing glands. Finally, lower BMP5 RNA levels were detected in patient breast tumors versus normal tissues, and correlated with cancer recurrence. Thus, the Ras effector Blimp-1 plays an essential role in TGF-β1--induced EMT via repression of BMP-5 in breast cancer. [ABSTRACT FROM AUTHOR]

The highly immunogenic human tumor antigen NY-ESO-1 (ESO) is a target of choice for anti-cancer immune therapy. In this study, we assessed spontaneous antibody (Ab) responses to ESO in a large cohort of patients with primary breast cancer (BC) and addressed the correlation between the presence of anti-ESO Ab, the expression of ESO in the tumors and their characteristics. We found detectable Ab responses to ESO in 1% of the patients. Tumors from patients with circulating Ab to ESO exhibited common characteristics, being mainly hormone receptor (HR)- invasive ductal carcinomas of high grade, including both HER2- and HER2+ tumors. In line with these results, we detected ESO expression in 20% of primary HR- BC, including both ESO Ab+ and Ab- patients, but not in HR+ BC. Interestingly, whereas expression levels in ESO+ BC were not significantly different between ESO Ab+ and Ab- patients, the former had, in average, significantly higher numbers of tumor-infiltrated lymph nodes, indicating that lymph node invasion may be required for the development of spontaneous anti-tumor immune responses. Thus, the presence of ESO Ab identifies a tumor subtype of HR- (HER2- or HER2+) primary BC with frequent ESO expression and, together with the assessment of antigen expression in the tumor, may be instrumental for the selection of patients for whom ESO-based immunotherapy may complement standard therapy. [ABSTRACT FROM AUTHOR]

Abstract: Interleukin-6 (IL-6) is a cytokine involved in different physiologic and pathophysiologic processes including carcinogenesis. In 2003, a single nucleotide polymorphism (−174G/C) of the IL-6 gene promoter has been linked to breast cancer prognosis in node-positive (N+) breast cancer patients. Since, different studies have led to conflicting conclusions about its role as a prognostic and/or diagnostic marker. The primary aim of our study was to investigate the link between −174G/C polymorphism and breast cancer risk on the one hand, and −174G/C polymorphism and prognosis in different groups of patients: sporadic N+breast cancers (n =138), sporadic N− breast cancers (n =95) and familial breast cancer (n =60) on the other hand. The variables of interest were disease-free survival and overall survival. The secondary aim of the study was to screen IL-6 gene promoter using direct sequencing to identify new polymorphisms in our French Caucasian breast cancer population. No association or trend of association between −174G/C polymorphism of IL-6 gene promoter gene and breast cancer diagnosis or prognosis was shown, even in meta-analyses. Furthermore, we have identified four novel polymorphic sites in the IL-6 gene promoter region: −764G→A, −757C→T, −233T→A, 15C→A. [Copyright &y& Elsevier]

Simple Summary: In the recent years, several deep learning methods for medical image segmentation have been developed for different purposes such as diagnosis, radiotherapy planning or to correlate images findings with other clinical data. However, few studies focus on longitudinal images and response assessment. To the best of our knowledge, this is the first study to date evaluating the use of automatic segmentation to obtain imaging biomarkers that can be used to assess treatment response in patients with metastatic breast cancer. Moreover, the statistical analysis of the different biomarkers shows that automatic segmentation can be successfully used for their computation, reaching similar performances compared to manual segmentation. Analysis also demonstrated the potential of the different biomarkers including novel/original ones to determine treatment response. Metastatic breast cancer patients receive lifelong medication and are regularly monitored for disease progression. The aim of this work was to (1) propose networks to segment breast cancer metastatic lesions on longitudinal whole-body PET/CT and (2) extract imaging biomarkers from the segmentations and evaluate their potential to determine treatment response. Baseline and follow-up PET/CT images of 60 patients from the EPICURE s e i n m e t a study were used to train two deep-learning models to segment breast cancer metastatic lesions: One for baseline images and one for follow-up images. From the automatic segmentations, four imaging biomarkers were computed and evaluated: SUL p e a k , Total Lesion Glycolysis (TLG), PET Bone Index (PBI) and PET Liver Index (PLI). The first network obtained a mean Dice score of 0.66 on baseline acquisitions. The second network obtained a mean Dice score of 0.58 on follow-up acquisitions. SUL p e a k , with a 32% decrease between baseline and follow-up, was the biomarker best able to assess patients' response (sensitivity 87%, specificity 87%), followed by TLG (43% decrease, sensitivity 73%, specificity 81%) and PBI (8% decrease, sensitivity 69%, specificity 69%). Our networks constitute promising tools for the automatic segmentation of lesions in patients with metastatic breast cancer allowing treatment response assessment with several biomarkers. [ABSTRACT FROM AUTHOR]

Background: The standard therapy regimen of conventional osteosarcoma includes neoadjuvant chemotherapy followed by surgical resection and postoperative chemotherapy. The percentage of necrotic tissue following induction chemotherapy is assessed by using the Huvos grading system, which classifies patients as "poor responders" (PR) and "good responders" (GR). The aim of this study was to identify molecular markers expressed differentially between good and poor responders to neoadjuvant chemotherapy in order to predict the response to chemotherapy in conventional osteosarcomas before beginning treatment.Methods: Suppression Substractive Hybridization (SSH) was performed by using cDNA from frozen biopsy specimens. Expression of selected relevant genes identified by SSH was validated by using QRT-PCR. Immunohistochemistry (IHC) on tissue microarray (TMA) sections of 52 biopsies was performed to investigate protein expression in an independent cohort.Results: ERK1 and STAT3 mRNA level were significantly different between PR and GR in an independent cohort. Phosphorylated STAT3 and ERK1 expressions by IHC on TMA were correlated with poor response to chemotherapy.Conclusions: Our results suggest that ERK1 and STAT3 expression are good predictive markers for chemotherapy response and that inhibitors might be used in combination with common chemotherapeutic drugs in conventional osteosarcomas. [ABSTRACT FROM AUTHOR]