Extreme Gradient Boosting Tuned with Metaheuristic Algorithms for Predicting Myeloid NGS Onco-Somatic Variant Pathogenicity.

The advent of next-generation sequencing (NGS) technologies has revolutionized the field of bioinformatics and genomics, particularly in the area of onco-somatic genetics. NGS has provided a wealth of information about the genetic changes that underlie cancer and has considerably improved our ability to diagnose and treat cancer. However, the large amount of data generated by NGS makes it difficult to interpret the variants. To address this, machine learning algorithms such as Extreme Gradient Boosting (XGBoost) have become increasingly important tools in the analysis of NGS data. In this paper, we present a machine learning tool that uses XGBoost to predict the pathogenicity of a mutation in the myeloid panel. We optimized the performance of XGBoost using metaheuristic algorithms and compared our predictions with the decisions of biologists and other prediction tools. The myeloid panel is a critical component in the diagnosis and treatment of myeloid neoplasms, and the sequencing of this panel allows for the identification of specific genetic mutations, enabling more accurate diagnoses and tailored treatment plans. We used datasets collected from our myeloid panel NGS analysis to train the XGBoost algorithm. It represents a data collection of 15,977 mutations variants composed of a collection of 13,221 Single Nucleotide Variants (SNVs), 73 Multiple Nucleoid Variants (MNVs), and 2683 insertion deletions (INDELs). The optimal XGBoost hyperparameters were found with Differential Evolution (DE), with an accuracy of 99.35%, precision of 98.70%, specificity of 98.71%, and sensitivity of 1.

Plasma nanoDSF Denaturation Profile at Baseline Is Predictive of Glioblastoma EGFR Status.

Glioblastoma (GBM) is the most frequent and aggressive primary brain tumor in adults. Recently, we demonstrated that plasma denaturation profiles of glioblastoma patients obtained using Differential Scanning Fluorimetry can be automatically distinguished from healthy controls with the help of Artificial Intelligence (AI). Here, we used a set of machine-learning algorithms to automatically classify plasma denaturation profiles of glioblastoma patients according to their EGFR status. We found that Adaboost AI is able to discriminate EGFR alterations in GBM with an 81.5% accuracy. Our study shows that the use of these plasma denaturation profiles could answer the unmet neuro-oncology need for diagnostic predictive biomarker in combination with brain MRI and clinical data, in order to allow for a rapid orientation of patients for a definitive pathological diagnosis and then treatment. We complete this study by showing that discriminating another mutation, MGMT, seems harder, and that post-surgery monitoring using our approach is not conclusive in the 48 h that follow the surgery.

Genomic analysis of paired IDHwt glioblastomas reveals recurrent alterations of MPDZ at relapse after radiotherapy and chemotherapy.

PURPOSE: We aimed to identify genomic drivers of glioblastoma inevitable recurrence. METHODS: Ten pairs of initial and recurrent frozen IDHwt glioblastoma samples were screened by CGH Array. Next Generation Sequencing (NGS) was then performed on an enriched cohort of 19 pairs. MPDZ alterations were analyzed using TCGA dataset. RESULTS: Nineteen IDHwt glioblastoma patients were included. Median age was 54.5 y/o (37.2-72.8). Using CGH array, unsupervised analysis aggregated the cohort by paired initial and recurrent tumors. Only 44% of CGH Array alterations were conserved at recurrence (amplifications: 55%; deletions: 30%). Two regions (including FPR1, 2 and 3) were lost at relapse: 19q13.33 and 19q13.41. MPDZ and 25 other genes were altered in ≥20% of recurrent tumors. NGS analysis of 29 candidate genes revealed 4 genes with pathogenic mutations: (FPR2, REL, TYRP1 and MPDZ). MPDZ (Multiple PDZ Domain Crumbs Cell Polarity Complex Component) was altered by two pathogenic mutations occurring at relapse. Using TCGA dataset we observed that a lower MPDZ mRNA expression was associated with IDHwt (p < 0.001) and grade IV (p < 0.001) gliomas. Finally, a low mRNA MPDZ expression was significantly correlated to poor overall survival in both IDHwt and IDH mutated gliomas, reinforcing the potential pejorative impact of MPDZ loss. CONCLUSION: Our results suggest that MPDZ is more frequently altered at relapse after radio-chemotherapy in glioblastoma IDHwt patients, suggesting that MPDZ impairment could contribute to the systematic resistance of these tumors opening new therapeutic perspectives.

Multiplexed Droplet Digital PCR Assays for the Simultaneous Screening of Major Genetic Alterations in Tumors of the Central Nervous System.

The increased integration of molecular alterations to define tumor type or grade in central nervous system (CNS) tumor classification brings new challenges for the pathologist to make the best use of a precious limited tissue specimen for molecular studies. Within the different methods available to identify gene alterations, the droplet digital PCR (dPCR) constitutes a rapid, cost-effective, and very sensitive tool. In this study, we describe the development and validation of five multiplexed dPCR assays to detect major CNS biomarkers by using only small amounts of DNA extracted from formalin-fixed paraffin-embedded specimens. When compared to HRM-sequencing, NGS-sequencing, RNA-sequencing, or simplex digital PCR assays used as "gold standard" methods, these multiplexed dPCR assays displayed 100% specificity and sensitivity for the simultaneous detection of: 1/BRAF V600E mutation and KIAA1549:BRAF fusion; 2/FGFR1 N546K and K656E mutations and FGFR1 duplication; 3/H3F3A K27M and G34R/V mutations; 4/IDH1 R132X and IDH2 R172X mutations; and 5/TERT promoter mutations C228T and C250T. In light of the increased integration of molecular alteration, we believe that such strategies might help laboratories to optimize their screening strategies for routine diagnosis of pediatric and adult CNS tumors.

Comparative genomic analysis of primary tumors and paired brain metastases in lung cancer patients by whole exome sequencing: a pilot study.

Lung cancer brain metastases (BMs) are frequent and associated with poor prognosis despite a better knowledge of lung cancer biology and the development of targeted therapies. The inconstant intracranial response to systemic treatments is partially due to tumor heterogeneity between the primary lung tumor (PLT) and BMs. There is therefore a need for a better understanding of lung cancer BMs biology to improve treatment strategies for these patients. We conducted a study of whole exome sequencing of paired BM and PLT samples. The number of somatic variants and chromosomal alterations was higher in BM samples. We identified recurrent mutations in BMs not found in PLT. Phylogenic trees and lollipop plots were designed to describe their functional impact. Among the 13 genes mutated in ≥ 1 BM, 7 were previously described to be associated with invasion process, including 3 with recurrent mutations in functional domains which may be future targets for therapy. We provide with some insights about the mechanisms leading to BMs. We found recurrent mutations in BM samples in 13 genes. Among these genes, 7 were previously described to be associated with cancer and 3 of them (CCDC178, RUNX1T1, MUC2) were described to be associated with the metastatic process.

Detection of EGFR, KRAS and BRAF mutations in metastatic cells from cerebrospinal fluid.

BACKGROUND: In lung adenocarcinoma, molecular profiling of actionable genes has become essential to set up targeted therapies. However, the feasibility and the relevance of molecular profiling from the cerebrospinal fluid (CSF) in the context of meningeal metastasis have been poorly assessed. METHODS: We selected patients with stage IV lung adenocarcinoma harbouring metastatic cells in the CSF after cytological analysis. Seven samples from six patients were eligible for molecular testing of epidermal growth factor receptor (EGFR), V-Ki-ras2 Kirsten rat sarcoma viral oncogene homologue (KRAS), v-Raf murine sarcoma viral oncogene homologue B1 (BRAF) and human epidermal growth factor receptor 2 (HER2) mutations using quantitative polymerase chain reaction (PCR) high-resolution melting curve analysis and Sanger sequencing after DNA extraction from the cell pellets of the CSF. RESULTS: Five patients showed mutations in one or two actionable genes, two harboured an EGFR mutation (exons 19 and 21), one only a KRAS mutation, one both EGFR and KRAS mutations and one a BRAF mutation. In all cases, the results of mutation testing provided new major information for patient management, leading to therapeutic adaptation. CSF molecular analysis identified mutations not detected in other neoplastic sites for two patients. In one case, the EGFR p.Thr790Met was identified. CSF was also the only sample available for genetic testing for almost all patients at the time of disease progression. CONCLUSIONS: When cancer cells are present in the CSF, the molecular profiling from the cell pellets is relevant, as it can detect supplemental or different mutations compared to a previous analysis of the primitive tumour or plasma cell-free DNA and allows the adaptation of the treatment strategy.

Droplet digital PCR is a powerful technique to demonstrate frequent FGFR1 duplication in dysembryoplastic neuroepithelial tumors.

Dysembryoplastic neuroepithelial tumors (DNT) share V600E mutation in the BRAF gene with other low grade neuroepithelial tumors (LGNTs). FGFR1 internal tandem duplication of the tyrosine-kinase domain (FGFR1-ITD), another genetic alteration that also leads to MAP kinase pathway alteration, has been previously reported in LGNTs by whole-genome sequencing. In the present study we searched for FGFR1-ITD by droplet digital PCR (DDPCR™) and for FGFR1 point mutations by HRM-sequencing in a series of formalin-fixed paraffin-embedded (FFPE) LGNTs including 12 DNT, 2 oligodendrogliomas lacking IDH mutation and 1p/19q co- deletion (pediatric-type oligodendrogliomas; PTOs), 3 pediatric diffuse astrocytomas (PDAs), 14 gangliogliomas (GGs) and 5 pilocytic astrocytomas (PAs). We showed by DDPCR™ that 5/12 DNT, but none of the other LGNTs, demonstrated FGFR1-ITD. In addition, these cases also accumulated phosphorylated-FGFR1 protein as shown by immunohistochemistry. FGFR1G539R point mutation was only recorded in one DNT that also showed FGFR1-ITD. Interestingly, these FGFR1 alterations were mutually exclusive from BRAFV600E mutation that was recorded in 13 LGNTs (3 DNTs, 1 PTO, 2 PDAs, 5 GGs and 2 PAs). Therefore, FGFR1 alteration mainly represented by FGFR1-ITD is a frequent event in DNT. DDPCR™ is an easy and alternative method than whole-genome sequencing to detect FGFR1-ITD in FFPE brain tumors, in routine practice.

Somatic mosaicism underlies X-linked acrogigantism syndrome in sporadic male subjects.

Somatic mosaicism has been implicated as a causative mechanism in a number of genetic and genomic disorders. X-linked acrogigantism (XLAG) syndrome is a recently characterized genomic form of pediatric gigantism due to aggressive pituitary tumors that is caused by submicroscopic chromosome Xq26.3 duplications that include GPR101 We studied XLAG syndrome patients (n= 18) to determine if somatic mosaicism contributed to the genomic pathophysiology. Eighteen subjects with XLAG syndrome caused by Xq26.3 duplications were identified using high-definition array comparative genomic hybridization (HD-aCGH). We noted that males with XLAG had a decreased log2ratio (LR) compared with expected values, suggesting potential mosaicism, whereas females showed no such decrease. Compared with familial male XLAG cases, sporadic males had more marked evidence for mosaicism, with levels of Xq26.3 duplication between 16.1 and 53.8%. These characteristics were replicated using a novel, personalized breakpoint junction-specific quantification droplet digital polymerase chain reaction (ddPCR) technique. Using a separate ddPCR technique, we studied the feasibility of identifying XLAG syndrome cases in a distinct patient population of 64 unrelated subjects with acromegaly/gigantism, and identified one female gigantism patient who had had increased copy number variation (CNV) threshold for GPR101 that was subsequently diagnosed as having XLAG syndrome on HD-aCGH. Employing a combination of HD-aCGH and novel ddPCR approaches, we have demonstrated, for the first time, that XLAG syndrome can be caused by variable degrees of somatic mosaicism for duplications at chromosome Xq26.3. Somatic mosaicism was shown to occur in sporadic males but not in females with XLAG syndrome, although the clinical characteristics of the disease were similarly severe in both sexes.

Differential expression of miR200a-3p and miR21 in grade II-III and grade IV gliomas: evidence that miR200a-3p is regulated by O6-methylguanine methyltransferase and promotes temozolomide responsiveness.

Glioblastoma multiforme (GBM) is the most common primary brain tumor and is among the deadliest of human cancers. Dysregulation of microRNAs (miRNAs) expression is an important step in tumor progression as miRNAs can act as tumor suppressors or oncogenes and may affect cell sensitivity to chemotherapy. Whereas the oncogenic miR21 has been shown to be overexpressed in gliomas, the expression and function of the tumor-supressor miR200a in GBMs remains unknown. In this study, we show that miR21 is upregulated in grade IV (GBMs) vs. grade II-III (LGs) gliomas, confirming that miR21 expression level is correlated with tumor grade, and that it may be considered as a marker of tumor progression. Conversely, miR200a is demonstrated for the first time to be downregulated in GBMs compared with LGs, and overexpression of miR200a in GBM cells is shown to promote TMZ-sensitivity. Interestingly, miR200a but not miR21 expression level is significantly higher in TMZ-responsive vs. -unresponsive tumoral glial cells in primary culture. Furthermore, miR200a appears negatively correlated with the expression of the DNA repair enzyme O (6)-methylguanine methyltransferase (MGMT), and the inhibition of MGMT activity results in an increase of miR200a expression in GBM cells. Taken together, these data strongly suggest that miR200a is likely to act as a crucial antitumoral factor regarding glioma progression. Interplay between miR200a and MGMT should be considered as potential mechanism involved in therapeutic response.

A multicenter blinded study evaluating EGFR and KRAS mutation testing methods in the clinical non-small cell lung cancer setting--IFCT/ERMETIC2 Project Part 1: Comparison of testing methods in 20 French molecular genetic National Cancer Institute platforms.

Epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors have limited use as first-line treatment for mutated EGFR metastatic non-small cell lung cancer. The French National Cancer Institute has installed molecular genetics platforms implementing EGFR and KRAS testing. However, there is considerable uncertainty as to which detection methods should be applied for routine diagnosis. This study aimed to compare the EGFR and KRAS genotyping methods developed by the IFCT/ERMETIC2 network platforms in two blind panels: 25 samples of serial dilutions of cell line DNA (20 centers) and 74 FFPE lung tumor samples (10 centers). The best threshold of mutation detection on cell lines was obtained using allele-specific amplification-based technologies. Nonamplifiable tissue samples were significantly less common when using alternative testing versus direct sequencing [15%; 95% confidence interval (CI), 14%-16% versus 40%; 95% CI, 39%-42%; P < 0.001]. Mutated cases increased from 42% (95% CI, 31%-54%) to 53% (95% CI, 41%-64%), with three supplementary EGFR mutations (p.G179A at exon 18 and p.L858R and p.L861Q at exon 21) and five supplementary KRAS mutations, when using alternative testing instead of direct sequencing. False-positive results were observed when using a PCR-based sizing assay, high-resolution melting, or pyrosequencing. Concordance analysis returned good kappa test scores for EGFR exon 19 and KRAS analysis when comparing sequencing with alternative methods and revealed no difference between alternative techniques themselves.

[Regional molecular genetics centers in thoracic oncology: what and who should be tested?]. / Plateformes de biologie moléculaire en oncologie thoracique: quelles recherches, pour qui?

Management of NSCLC patients is more and more individualized especially on the base of bioguided treatments. In order to guarantee an access for all the patients too this type of strategy, the French NCI supports since 2006 a nationwide network of 28 regional genetics center. The financial support is based on public funds. The French NCI recommends today the assessment of seven biomarkers for all stage IV non squamous NSCLC patients. Due to financial and technical reasons, this recommendation must be followed. However, the molecular profiling of lung cancer patients would ideally be extended across all stages and all histological types of the disease in order to improve our knowledge in this field and provides the patient with an opportunity to access a bioguided treatment as frequently as possible.

Dysembryoplastic neuroepithelial tumors share with pleomorphic xanthoastrocytomas and gangliogliomas BRAF(V600E) mutation and expression.

Pediatric cortical glioneuronal benign tumors mainly include gangliogliomas (GG) [differential diagnoses pilocytic astrocytomas (PA) and pleomorphic xanthoastrocytomas (PXA)] and dysembryoplastic neuroepithelial tumor (DNT). DNT include the specific form and the controversial non-specific form that lack the specific glioneuronal element. Our aims were to search for BRAF(V600E) mutation and CD34 expression in DNT, PXA, GG and PA to correlate BRAF(V600E) mutation with BRAF(V600E) expression and to evaluate their diagnostic and prognostic values. Ninety-six children were included. BRAF(V600E) mutation was studied by sequencing and immunohistochemistry; CD34 expression was analyzed by immunohistochemistry. BRAF(V600E) mutation was detected in PXA (60%), GG (38.7%), DNT (30%, including 3/11 specific and 3/9 non-specific forms) and PA (12.5%). BRAF(V600E) expression was recorded in PXA (60%), GG (45.2%) and DNT (30%). CD34 expression was recorded in PXA (60%), GG (58.1%), DNT (25%) and PA (12.5%). Neither CD34 expression nor BRAF(V600E) status was predictive of prognosis, except for PA tumors where CD34 expression was associated with a shorter overall survival. In conclusion, DNT shared with PXA and GG, BRAF(V600E) mutation and/or CD34 expression, which represent molecular markers for these tumors, and we recommend searching for CD34 expression and BRAF(V600E) mutation in all DNT, especially the non-specific forms.

IDH mutation status impact on in vivo hypoxia biomarkers expression: new insights from a clinical, nuclear imaging and immunohistochemical study in 33 glioma patients.

Mutations in the gene encoding isocitrate dehydrogenase enzyme isoforms 1 (IDH1) and 2 (IDH2) have recently been identified in a large proportion of glial tumors of the CNS, but their mechanistic role in tumor development remains unclear. Here, we assessed the actual impact of IDH1 and IDH2 mutations in patients harboring WHO grade II and III gliomas. We sequenced IDH1 at codon 132 and IDH2 at codon 172 in 33 patients with WHO grade II and III gliomas who benefited from a preoperative (18)F-FDG positron emission tomography (PET). Immunohistochemical expression of Hypoxia Inducible Factor-1alpha (HIF-1α), Carbonic Anhydrase IX (CAIX), Glucose Transporter 1 (GLUT1) and Caspase 3 active form (CASP3) along with the R132HIDH1 mutation was assessed in all cases as well as 1p/19q deletion status and p53 expression. HIF-1α expression was found in 15% of IDH-mutated compared to 7.7% of IDH-nonmutated tumors (P = 0.954). Also, GLUT-1 positive staining was found in 5% of IDH-mutated and in 7.1% of IDH-nonmutated tumors (P = 0.794). Finally, CA-IX expression was found in 15% of IDH-mutated and in 7.7% of IDH-nonmutated tumors (P = 0.484). The combined expression of these three hypoxic markers was found in two WHO grade III tumors, one of which was IDH-mutated whereas the other was IDH-nonmutated (P = 0.794). In IDH-mutated tumors, the median SUVmax ratio was 2.24 versus 2.15 in IDH-nonmutated tumors (P = 0.775). Together, these data question the actual relationship between IDH mutation status and in vivo hypoxic biomarkers expression in WHO grade II and III gliomas.

Prognostic impact of CD133 mRNA expression in 48 glioblastoma patients treated with concomitant radiochemotherapy: a prospective patient cohort at a single institution.

BACKGROUND: Cancer stem cells are thought to represent the population of tumorigenic cells responsible for tumor development. The CD133 antigen has been described as a putative stem cell marker in malignant brain tumor that could identify such a tumorigenic population in a subset of glioblastoma. To date, the correlation between CD133 expression in primary glioblastoma and patient prognosis is not clearly established. To address this question we investigated the relationship between CD133 mRNA expression and patient outcome in a glioblastoma patient cohort. MATERIALS AND METHODS: The quantitative expression of CD133 stem cell antigen mRNA using real-time QRT-PCR was assessed in a cohort of 48 consecutive primary glioblastoma patients treated by chemoradiation with temozolomide. RESULTS: On multivariate survival analysis, high CD133 mRNA expression was a significant (P = 0.007) prognostic factor for adverse progression-free and overall survival independent of extent of resection (P = 0.012) and MGMT methylation status (P = 0.002). Patient age was also an independent prognosticator of overall survival (P = 0.037). Furthermore, according to the conjoined expression of CD133 mRNA and MGMT status, the patients were categorized into 3 groups with homogenous prognosis. CONCLUSIONS: These findings constitute conclusive evidence that the measurement of the mRNA expression of CD133 stem cell antigen actually impacts the survival of GBM patients.

Adrenomedullin expression and regulation in human glioblastoma, cultured human glioblastoma cell lines and pilocytic astrocytoma.

Clinical and experimental studies suggest that angiogenesis is a prerequisite for solid tumour growth. Glioblastoma (GBM) and pilocytic astrocytoma (PA), both angiogenic tumours display strong contrast enhancement associated with peripheral oedema in GBM but not in PA indicating differences in vascular permeability in these two types of gliomas. Here we show that expression of adrenomedullin (AM) mRNA is induced in GBM whereas is barely detectable in PA. In situ analysis of tumour specimens undergoing neovascularisation shows that the production of AM is specifically induced in a subset of GBM cells distinguished by their immediate proximity to necrotic foci (presumably hypoxic regions), suggesting a hypoxic induction of AM expression in GBM. Vascular endothelial growth factor (VEGF) mRNA levels are increased in GBM and moderate in PA. Immunohistochemical study showed that cytoplasmic AM, VEGF and HIF-1α nuclear immunoreactivity were recorded in GBM located near large necrotic areas whereas they were not expressed by PA tumour cells. Interestingly, double fluorescence immunostaining demonstrated that 85% of AM immunoreactivity colocalised with VEGF. AM transduces its effects through calcitonin receptor-like receptor/receptor activity modifying protein-2 and -3 (CLR/RAMP2 and CLR/RAMP3). Real-time quantitative RT-PCR showed expression of RAMP2, RAMP3 and CLR in PA and GBM, suggesting that AM may function as an autocrine/paracrine growth factor for GBM cells. These observations strongly support the concept that tumour angiogenesis is regulated by paracrine mechanisms and identify beside VEGF, AM as a potential tumour angiogenesis factor in vivo which constitutes a potential interesting molecular target in GBM treatment.

Non-small cell lung cancer-smokers or non-smokers, does it matter?

Besides epidermal growth factor receptor (EGFR) gene mutations, clinical factors such as smoking status have been identified as predictors for survival for NSCLC patients treated with EGFR-tyrosine kinase inhibitors (TKI). However, the biological screening for EGFR gene mutations is not routinely available everywhere. Therefore, the question arises if the decision to treat patients with EGFR-TKI should be based on clinical factors, and in particular smoking status, alone. We illustrate the difficulties faced by clinicians with the case of a 56-year-old man with stage IV lung adenocarcinoma and a smoking history of 30-pack-year. This patient received erlotinib first-line after its enrolment in a clinical trial. After 4 months, he presented with a dramatic clinical and radiological response. The biological analysis of the tumour revealed an EGFR exon 19 deletion. This report emphasizes that smoking status alone appears inappropriate in selecting patients for EGFR-TKI treatment. In addition, the relatively high number of (ex-)smokers retrieved from prospective studies on NSCLC patients with tumours showing an EGFR mutation should be emphasized, as it represents up to 30% of patients. Therefore, a biological rather than a clinical selection for patients' eligibility for EGFR-TKI appears mandatory.