Cancer mutational signatures identification in clinical assays using neural embedding-based representations.

While mutational signatures provide a plethora of prognostic and therapeutic insights, their application in clinical-setting, targeted gene panels is extremely limited. We develop a mutational representation model (which learns and embeds specific mutation signature connections) that enables prediction of dominant signatures with only a few mutations. We predict the dominant signatures across more than 60,000 tumors with gene panels, delineating their landscape across different cancers. Dominant signature predictions in gene panels are of clinical importance. These included UV, tobacco, and apolipoprotein B mRNA editing enzyme, catalytic polypeptide (APOBEC) signatures that are associated with better survival, independently from mutational burden. Further analyses reveal gene and mutation associations with signatures, such as SBS5 with TP53 and APOBEC with FGFR3S249C. In a clinical use case, APOBEC signature is a robust and specific predictor for resistance to epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKIs). Our model provides an easy-to-use way to detect signatures in clinical setting assays with many possible clinical implications for an unprecedented number of cancer patients.

Multimodal Machine Learning for Prediction of 30-Day Readmission Risk in Elderly Population.

BACKGROUND: Readmission within 30 days is a prevalent issue among elderly patients, linked to unfavorable health outcomes. Our objective was to develop and validate multimodal machine learning models for predicting 30-day readmission risk in elderly patients discharged from internal medicine departments. METHODS: This was a retrospective cohort study which included elderly patients aged 75 or older, who were hospitalized at the Hadassah Medical Center internal medicine departments between 2014 and 2020. Three machine learning algorithms were developed and employed to predict 30-day readmission risk. The primary measures were predictive model performance scores, specifically area under the receiver operator curve (AUROC), and average precision. RESULTS: This study included 19,569 admissions. Of them, 3258 (16.65%) resulted in 30-day readmission. Our 3 proposed models demonstrated high accuracy and precision on an unseen test set, with AUROC values of 0.87, 0.89, and 0.93, respectively, and average precision values of 0.76, 0.78, and 0.81. Feature importance analysis revealed that the number of admissions in the past year, history of 30-day readmission, Charlson score, and admission length were the most influential variables. Notably, the natural language processing score, representing the probability of readmission according to a textual-based model trained on social workers' assessment letters during hospitalization, ranked among the top 10 contributing factors. CONCLUSIONS: Leveraging multimodal machine learning offers a promising strategy for identifying elderly patients who are at high risk for 30-day readmission. By identifying these patients, machine learning models may facilitate the effective execution of preventive actions to reduce avoidable readmission incidents.

Both cell autonomous and non-autonomous processes modulate the association between replication timing and mutation rate.

Cancer somatic mutations are the product of multiple mutational and repair processes, some of which are tightly associated with DNA replication. Mutation rates (MR) are known to be higher in late replication timing (RT) regions, but different processes can affect this association. Systematic analysis of the mutational landscape of 2787 tumors from 32 tumor types revealed that approximately one third of the tumor samples show weak association between replication timing and mutation rate. Further analyses revealed that those samples have unique mutational signatures and are enriched with mutations in genes involved in DNA replication, DNA repair and chromatin structure. Surprisingly, analysis of differentially expressed genes between weak and strong RT-MR association groups revealed that tumors with weak association are enriched with genes associated with cell-cell communication and the immune system, suggesting a non-autonomous response to DNA damage.

Mutational signatures association with replication timing in normal cells reveals similarities and differences with matched cancer tissues.

Mutational signatures' association with replication timing (RT) has been studied in cancer samples, but the RT distribution of somatic mutations in non-cancerous cells was only minimally explored. Here, we performed comprehensive analyses of mutational signatures in 2.9 million somatic mutations across multiple non-cancerous tissues, stratified by early and late RT regions. We found that many mutational processes are active mainly or solely in early RT, such as SBS16 in hepatocytes and SBS88 in the colon, or in late RT, such as SBS4 in lung and hepatocytes, and SBS18 across many tissues. The two ubiquitous signatures, SBS1 and SBS5, showed late and early bias, respectively, across multiple tissues and in mutations representing germ cells. We also performed a direct comparison with cancer samples in 4 matched tissue-cancer types. Unexpectedly, while for most signatures the RT bias was consistent in normal tissue and in cancer, we found that SBS1's late RT bias is lost in cancer.

Distinct breast cancer phenotypes in BRCA 1/2 carriers based on ER status.

PURPOSE: BRCA1/2 genes are the two main genes associated with hereditary breast cancers (BC). In the present study, we explore clinical and molecular characteristics of BRCA-associated BC in relation to estrogen receptor (ER) status. METHODS: Three BC databases (DB) were evaluated: (i) Hadassah oncogenetics (n = 4826); (ii) METABRIC (n = 1980), and (iii) Nick-Zainal (n = 560). We evaluated age at diagnosis in BRCA positive (BP) and BRCA negative (BN) patients, and tested for mutational signature differences in cohort iii. mRNA differential expression analysis (DEA) and pathway analysis were performed in cohort ii. RESULTS: Age at diagnosis was lower in BP vs. BN tumors in all cohorts in the ER- group, and only in cohort i for the ER + group. Signature 3 was universal in BP BC, whereas several signatures were associated with ER status. Pathway analysis was performed between BP&BN, and was significant only in ER- tumors: the major activated pathways involved cancer-related processes and were highly significant. The most significant pathway was estrogen-mediated S-phase entry and the most activated upstream regulator was ERBB2. CONCLUSION: Signature 3 was universal for all BP BC, while other signatures were associated with ER status. ER + BP& BN show similar genomic characteristics, ER- BP differed markedly from BN. This suggests that the initial carcinogenic process is universal for all BRCA carriers, but further insults lead to the development of two genomically distinct subtypes ER- and ER + . This may shed light on possible mechanisms involved in BP and carry preventive and therapeutic implications.

Concordance between cancer gene alterations in tumor and circulating tumor DNA correlates with poor survival in a real-world precision-medicine population.

Genomic analysis, performed on tumoral tissue DNA and on circulating tumor DNA (ctDNA) from blood, is the cornerstone of precision cancer medicine. Herein, we characterized the clinical prognostic implications of the concordance of alterations in major cancer genes between tissue- and blood-derived DNA in a pan-cancer cohort. The molecular profiles of both liquid (Guardant Health) and tissue (Foundation Medicine) biopsies from 433 patients were analyzed. Mutations and amplifications of cancer genes scored by these two tests were assessed. In 184 (42.5%) patients, there was at least one mutual gene alteration. The mean number of mutual gene-level alterations in the samples was 0.67 per patient (range: 0-5). A higher mutual gene-level alteration number correlated with shorter overall survival (OS). As confirmed in multivariable analysis, patients with ≥2 mutual gene-level alterations in blood and tissue had a hazard ratio (HR) of death of 1.49 (95% confidence interval [CI]=1-2.2; P=0.047), whereas patients with ≥3 mutual gene-level alterations had an HR of death 2.38 (95% CI=1.47-3.87; P=0.0005). Together, our results show that gene-level concordance between tissue DNA and ctDNA analysis is prevalent and is an independent factor predicting significantly shorter patient survival.

Shared Cancer Dataset Analysis Identifies and Predicts the Quantitative Effects of Pan-Cancer Somatic Driver Variants.

Driver mutations endow tumors with selective advantages and produce an array of pathogenic effects. Determining the function of somatic variants is important for understanding cancer biology and identifying optimal therapies. Here, we compiled a shared dataset from several cancer genomic databases. Two measures were applied to 535 cancer genes based on observed and expected frequencies of driver variants as derived from cancer-specific rates of somatic mutagenesis. The first measure comprised a binary classifier based on a binomial test; the second was tumor variant amplitude (TVA), a continuous measure representing the selective advantage of individual variants. TVA outperformed all other computational tools in terms of its correlation with experimentally derived functional scores of cancer mutations. TVA also highly correlated with drug response, overall survival, and other clinical implications in relevant cancer genes. This study demonstrates how a selective advantage measure based on a large cancer dataset significantly impacts our understanding of the spectral effect of driver variants in cancer. The impact of this information will increase as cancer treatment becomes more precise and personalized to tumor-specific mutations. SIGNIFICANCE: A new selective advantage estimation assists in oncogenic driver identification and relative effect measurements, enabling better prognostication, therapy selection, and prioritization.

Identification of growth hormone receptor as a relevant target for precision medicine in low-EGFR expressing glioblastoma.

OBJECTIVE: New therapeutic approaches are needed to improve the prognosis of glioblastoma (GBM) patients. METHODS: With the objective of identifying alternative oncogenic mechanisms to abnormally activated epidermal growth factor receptor (EGFR) signalling, one of the most common oncogenic mechanisms in GBM, we performed a comparative analysis of gene expression profiles in a series of 54 human GBM samples. We then conducted gain of function as well as genetic and pharmocological inhibition assays in GBM patient-derived cell lines to functionnally validate our finding. RESULTS: We identified that growth hormone receptor (GHR) signalling defines a distinct molecular subset of GBMs devoid of EGFR overexpression. GHR overexpression was detected in one third of patients and was associated with low levels of suppressor of cytokine signalling 2 (SOCS2) expression due to SOCS2 promoter hypermethylation. In GBM patient-derived cell lines, GHR signalling modulates the expression of proteins involved in cellular movement, promotes cell migration, invasion and proliferation in vitro and promotes tumourigenesis, tumour growth, and tumour invasion in vivo. GHR genetic and pharmacological inhibition reduced cell proliferation and migration in vitro. CONCLUSION: This study pioneers a new field of investigation to improve the prognosis of GBM patients.

High throughput SARS-CoV-2 variant analysis using molecular barcodes coupled with next generation sequencing.

The identification of SARS-CoV-2 variants across the globe and their implications on the outspread of the pandemic, infection potential and resistance to vaccination, requires modification of the current diagnostic methods to map out viral mutations rapidly and reliably. Here, we demonstrate that integrating DNA barcoding technology, sample pooling and Next Generation Sequencing (NGS) provide an applicable solution for large-population viral screening combined with specific variant analysis. Our solution allows high throughput testing by barcoding each sample, followed by pooling of test samples using a multi-step procedure. First, patient-specific barcodes are added to the primers used in a one-step RT-PCR reaction, amplifying three different viral genes and one human housekeeping gene (as internal control). Then, samples are pooled, purified and finally, the generated sequences are read using an Illumina NGS system to identify the positive samples with a sensitivity of 82.5% and a specificity of 97.3%. Using this solution, we were able to identify six known and one unknown SARS-CoV-2 variants in a screen of 960 samples out of which 258 (27%) were positive for the virus. Thus, our diagnostic solution integrates the benefits of large population and epidemiological screening together with sensitive and specific identification of positive samples including variant analysis at a single nucleotide resolution.

Rituximab, methotrexate, procarbazine and lomustine (R-MPL) for the treatment of primary Central nervous system lymphoma.

The optimal high-dose methotrexate (HDMTX)-based combination therapy for primary central nervous system lymphoma is unknown. We report our experience with rituximab, HDMTX, procarbazine and lomustine (R-MPL) given as first-line treatment in our center. Fifty-two patients between 2006 and 2019 were included. Eighteen patients proceeded to autologous transplant or two cycles of intermediate-dose cytarabine. The median age was 62 y (range 28-94) and the Eastern Cooperative Oncology Group performance status (ECOG-PS) was ≥2 in 62% (32/52). The overall/complete response rates were 79% (41/52) and 52% (27/52), respectively. The median progression-free/overall survival was 19 m/84m, respectively. Grade 3-4 adverse events included infections (17%) and kidney injury (13%). Ten patients (19%) discontinued therapy for toxicity. There were no treatment-related deaths. In summary, in a cohort enriched in frail patients, R-MPL achieved good responses and OS and was safe for all ages. The PFS was sub-optimal, possibly explained by a low proportion of consolidation. This regimen should be evaluated prospectively.

TP53_PROF: a machine learning model to predict impact of missense mutations in TP53.

Correctly identifying the true driver mutations in a patient's tumor is a major challenge in precision oncology. Most efforts address frequent mutations, leaving medium- and low-frequency variants mostly unaddressed. For TP53, this identification is crucial for both somatic and germline mutations, with the latter associated with the Li-Fraumeni syndrome (LFS), a multiorgan cancer predisposition. We present TP53_PROF (prediction of functionality), a gene specific machine learning model to predict the functional consequences of every possible missense mutation in TP53, integrating human cell- and yeast-based functional assays scores along with computational scores. Variants were labeled for the training set using well-defined criteria of prevalence in four cancer genomics databases. The model's predictions provided accuracy of 96.5%. They were validated experimentally, and were compared to population data, LFS datasets, ClinVar annotations and to TCGA survival data. Very high accuracy was shown through all methods of validation. TP53_PROF allows accurate classification of TP53 missense mutations applicable for clinical practice. Our gene specific approach integrated machine learning, highly reliable features and biological knowledge, to create an unprecedented, thoroughly validated and clinically oriented classification model. This approach currently addresses TP53 mutations and will be applied in the future to other important cancer genes.

Cancer Mutational Processes Vary in Their Association with Replication Timing and Chromatin Accessibility.

Cancer somatic mutations are the product of multiple mutational and repair processes, both of which are tightly associated with DNA replication. Distinctive patterns of somatic mutation accumulation, termed mutational signatures, are indicative of processes sustained within tumors. However, the association of various mutational processes with replication timing (RT) remains an open question. In this study, we systematically analyzed the mutational landscape of 2,787 tumors from 32 tumor types separately for early and late replicating regions using sequence context normalization and chromatin data to account for sequence and chromatin accessibility differences. To account for sequence differences between various genomic regions, an artificial genome-based approach was developed to expand the signature analyses to doublet base substitutions and small insertions and deletions. The association of mutational processes and RT was signature specific: Some signatures were associated with early or late replication (such as SBS7b and SBS7a, respectively), and others had no association. Most associations existed even after normalizing for genome accessibility. A focused mutational signature identification approach was also developed that uses RT information to improve signature identification; this approach found that SBS16, which is biased toward early replication, is strongly associated with better survival rates in liver cancer. Overall, this novel and comprehensive approach provides a better understanding of the etiology of mutational signatures, which may lead to improved cancer prevention, diagnosis, and treatment. SIGNIFICANCE: Many mutational processes associate with early or late replication timing regions independently of chromatin accessibility, enabling development of a focused identification approach to improve mutational signature detection.

Comprehensive assessment of TP53 loss of function using multiple combinatorial mutagenesis libraries.

The diagnosis of somatic and germline TP53 mutations in human tumors or in individuals prone to various types of cancer has now reached the clinic. To increase the accuracy of the prediction of TP53 variant pathogenicity, we gathered functional data from three independent large-scale saturation mutagenesis screening studies with experimental data for more than 10,000 TP53 variants performed in different settings (yeast or mammalian) and with different readouts (transcription, growth arrest or apoptosis). Correlation analysis and multidimensional scaling showed excellent agreement between all these variables. Furthermore, we found that some missense mutations localized in TP53 exons led to impaired TP53 splicing as shown by an analysis of the TP53 expression data from the cancer genome atlas. With the increasing availability of genomic, transcriptomic and proteomic data, it is essential to employ both protein and RNA prediction to accurately define variant pathogenicity.

Survival Implications of the Relationship between Tissue versus Circulating Tumor DNA TP53 Mutations-A Perspective from a Real-World Precision Medicine Cohort.

Interrogating the genomics of circulating tumor DNA (ctDNA; the liquid biopsy) has advantages in patients in whom tissue biopsy is difficult. However, the reported concordance between genomic analysis of tissue DNA and ctDNA is variable among studies. Herein, we characterized the clinical implications of the relationship between mutations in TP53 genes in tissue DNA versus ctDNA. The molecular profiles of both liquid (Guardant Health) and tissue (Foundation Medicine) biopsies from 433 patients were analyzed (pan-cancer setting). In 71/433 (16%) cases, all same TP53 mutations were detected in both tissue DNA and ctDNA; in 18/433 (4%), same mutation plus additional mutation/mutations; and in 27/433 (6%), different TP53 mutations were detected. In 99/433 (23%) cases, TP53 mutations were detected only in tissue DNA; in 43/433 (10%), only in ctDNA; and in 175/433 (40%), no TP53 mutations were detected in either test. When TP53 mutations were identical in tissue and ctDNA, the alterations were enriched for nonsense mutations, and survival was significantly shorter in multivariate analysis (as compared with different mutations in ctDNA vs. tissue or no mutations); this finding was independent of tumor type, time interval between tests, and the %ctDNA for TP53 mutations. In summary, in 16% of 433 patients with diverse cancers, TP53 mutations were identical in tissue DNA and ctDNA. In these individuals, the alterations were enriched for stop-gain (nonsense) mutations (results in a premature termination codon). Though unknown confounders cannot be ruled out, these patients fared significantly worse than those whose ctDNA and tissue DNA harbored different TP53 mutation portfolios or no TP53 mutations.

High prevalence of cancer-associated TP53 variants in the gnomAD database: A word of caution concerning the use of variant filtering.

The 1,000 genome project, the Exome Aggregation Consortium (ExAC) or the Genome Aggregation database (gnomAD) datasets, were developed to provide large-scale reference data of genetic variations for various populations to filter out common benign variants and identify rare variants of clinical importance based on their frequency in the human population. Using a TP53 repository of 80,000 cancer variants, as well as TP53 variants from multiple cancer genome projects, we have defined a set of certified oncogenic TP53 variants. This specific set has been independently validated by functional and in silico predictive analysis. Here we show that a significant number of these variants are included in gnomAD and ExAC. Most of them correspond to TP53 hotspot variants occurring as somatic and germline events in human cancer. Similarly, disease-associated variants for five other tumor suppressor genes, including BRCA1, BRCA2, APC, PTEN, and MLH1, have also been identified. This study demonstrates that germline TP53 variants in the human population are more frequent than previously thought. Furthermore, population databases such as gnomAD or ExAC must be used with caution and need to be annotated for the presence of oncogenic variants to improve their clinical utility.

Machine Learning for Better Prognostic Stratification and Driver Gene Identification Using Somatic Copy Number Variations in Anaplastic Oligodendroglioma.

BACKGROUND: 1p/19q-codeleted anaplastic gliomas have variable clinical behavior. We have recently shown that the common 9p21.3 allelic loss is an independent prognostic factor in this tumor type. The aim of this study is to identify less frequent genomic copy number variations (CNVs) with clinical importance that may shed light on molecular oncogenesis of this tumor type. MATERIALS AND METHODS: A cohort of 197 patients with anaplastic oligodendroglioma was collected as part of the French POLA network. Clinical, pathological, and molecular information was recorded. CNV analysis was performed using single-nucleotide polymorphism arrays. Computational biology and feature selection based on the random forests method were used to identify CNV events associated with overall survival and other clinical-pathological variables. RESULTS: Recurrent chromosomal events were identified in chromosomes 4, 9, and 11. Forty-six focal amplification events and 22 focal deletion events were identified. Twenty-four focal CNV areas were associated with survival, and five of them were significantly associated with survival after multivariable analysis. Nine out of 24 CNV events were validated using an external cohort of The Cancer Genome Atlas. Five of the validated events contain a cancer-related gene or microRNA: CDKN2A deletion, SS18L1 amplification, RHOA/MIR191 copy-neutral loss of heterozygosity, FGFR3 amplification, and ARNT amplification. The CNV profile contributes to better survival prediction compared with clinical-based risk assessment. CONCLUSION: Several recurrent CNV events, detected in anaplastic oligodendroglioma, enable better survival prediction. More importantly, they help in identifying potential genes for understanding oncogenesis and for personalized therapy. IMPLICATIONS FOR PRACTICE: Genomic analysis of 197 anaplastic oligodendroglioma tumors reveals recurrent somatic copy number variation areas that may help in understanding oncogenesis and target identification for precision medicine. A machine learning multivariable model built using this genomic information enables better survival prediction.

A recurrent point mutation in PRKCA is a hallmark of chordoid gliomas.

Chordoid glioma (ChG) is a characteristic, slow growing, and well-circumscribed diencephalic tumor, whose mutational landscape is unknown. Here we report the analysis of 16 ChG by whole-exome and RNA-sequencing. We found that 15 ChG harbor the same PRKCA D463H mutation. PRKCA encodes the Protein kinase C (PKC) isozyme alpha (PKCα) and is mutated in a wide range of human cancers. However the hot spot PRKCA D463H mutation was not described in other tumors. PRKCA D463H is strongly associated with the activation of protein translation initiation (EIF2) pathway. PKCαD463H mRNA levels are more abundant than wild-type PKCα transcripts, while PKCαD463H is less stable than the PCKαWT protein. Compared to PCKαWT, the PKCαD463H protein is depleted from the cell membrane. The PKCαD463H mutant enhances proliferation of astrocytes and tanycytes, the cells of origin of ChG. In conclusion, our study identifies the hallmark mutation for chordoid gliomas and provides mechanistic insights on ChG oncogenesis.

Correction to: ATP binding cassette (ABC) transporters: expression and clinical value in glioblastoma.

The names of authors Marc Sanson and Jean-Yves Delattre were incorrectly presented in the initial online publication. The original article has been corrected.

ATP binding cassette (ABC) transporters: expression and clinical value in glioblastoma.

ATP-binding cassette transporters (ABC transporters) regulate traffic of multiple compounds, including chemotherapeutic agents, through biological membranes. They are expressed by multiple cell types and have been implicated in the drug resistance of some cancer cells. Despite significant research in ABC transporters in the context of many diseases, little is known about their expression and clinical value in glioblastoma (GBM). We analyzed expression of 49 ABC transporters in both commercial and patient-derived GBM cell lines as well as from 51 human GBM tumor biopsies. Using The Cancer Genome Atlas (TCGA) cohort as a training dataset and our cohort as a validation dataset, we also investigated the prognostic value of these ABC transporters in newly diagnosed GBM patients, treated with the standard of care. In contrast to commercial GBM cell lines, GBM-patient derived cell lines (PDCL), grown as neurospheres in a serum-free medium, express ABC transporters similarly to parental tumors. Serum appeared to slightly increase resistance to temozolomide correlating with a tendency for an increased expression of ABCB1. Some differences were observed mainly due to expression of ABC transporters by microenvironmental cells. Together, our data suggest that the efficacy of chemotherapeutic agents may be misestimated in vitro if they are the targets of efflux pumps whose expression can be modulated by serum. Interestingly, several ABC transporters have prognostic value in the TCGA dataset. In our cohort of 51 GBM patients treated with radiation therapy with concurrent and adjuvant temozolomide, ABCA13 overexpression is associated with a decreased progression free survival in univariate (p < 0.01) and multivariate analyses including MGMT promoter methylation (p = 0.05) suggesting reduced sensitivity to temozolomide in ABCA13 overexpressing GBM. Expression of ABC transporters is: (i) detected in GBM and microenvironmental cells and (ii) better reproduced in GBM-PDCL. ABCA13 expression is an independent prognostic factor in newly diagnosed GBM patients. Further prospective studies are warranted to investigate whether ABCA13 expression can be used to further personalize treatments for GBM.

Ex-vivo organ culture as potential prioritization tool for breast cancer targeted therapy.

The growing use of genomic testing presents new treatment options but also new dilemmas. We describe here a heavily-pretreated metastatic triple negative breast cancer patient who failed to respond to conventional treatment. Genomic analyses were performed that discovered several targetable alterations (e.g. FGFR1, CDK6, INSR) and created a clinical challenge - which target to target first? Our solution to this relatively common scenario was using ex-vivo organ culture (EVOC) system to prioritize treatment directed toward the best molecular target. EVOC enabled the trial of several potent targeted agents (Everolimus, Linsitinib, Palbociclib, AZD4547) and allowed semi-quantitative measurement of tumor response. The best response was to FGFR inhibitor, AZD4547. Consequently, the most accessible FGFR inhibiting agents (Pazopanib, then Nintedanib) were administered and some response was achieved. This report provides a potential rationale for utilizing EVOC system to predict tumor response to targeted therapy when multiple targets are proposed.