Genomic portrait of resectable hepatocellular carcinomas: implications of RB1 and FGF19 aberrations for patient stratification.

UNLABELLED: Hepatic resection is the most curative treatment option for early-stage hepatocellular carcinoma, but is associated with a high recurrence rate, which exceeds 50% at 5 years after surgery. Understanding the genetic basis of hepatocellular carcinoma at surgically curable stages may enable the identification of new molecular biomarkers that accurately identify patients in need of additional early therapeutic interventions. Whole exome sequencing and copy number analysis was performed on 231 hepatocellular carcinomas (72% with hepatitis B viral infection) that were classified as early-stage hepatocellular carcinomas, candidates for surgical resection. Recurrent mutations were validated by Sanger sequencing. Unsupervised genomic analyses identified an association between specific genetic aberrations and postoperative clinical outcomes. Recurrent somatic mutations were identified in nine genes, including TP53, CTNNB1, AXIN1, RPS6KA3, and RB1. Recurrent homozygous deletions in FAM123A, RB1, and CDKN2A, and high-copy amplifications in MYC, RSPO2, CCND1, and FGF19 were detected. Pathway analyses of these genes revealed aberrations in the p53, Wnt, PIK3/Ras, cell cycle, and chromatin remodeling pathways. RB1 mutations were significantly associated with cancer-specific and recurrence-free survival after resection (multivariate P = 0.038 and P = 0.012, respectively). FGF19 amplifications, known to activate Wnt signaling, were mutually exclusive with CTNNB1 and AXIN1 mutations, and significantly associated with cirrhosis (P = 0.017). CONCLUSION: RB1 mutations can be used as a prognostic molecular biomarker for resectable hepatocellular carcinoma. Further study is required to investigate the potential role of FGF19 amplification in driving hepatocarcinogenesis in patients with liver cirrhosis and to investigate the potential of anti-FGF19 treatment in these patients.

Learning Polymorphic Neural ODEs With Time-Evolving Mixture.

Neural ordinary differential equations (NODE) present a new way of considering a deep residual network as a continuous structure by layer depth. However, it fails to overcome its representational limits, where it cannot learn all possible homeomorphisms of input data space, and therefore quickly saturates in terms of performance even as the number of layers increases. Here, we show that simply stacking Neural ODE blocks could easily improve performance by alleviating this issue. Furthermore, we suggest a more effective way of training neural ODE by using a time-evolving mixture weight on multiple ODE functions that also evolves with a separate neural ODE. We provide empirical results that are suggestive of improved performance over stacked as well as vanilla neural ODEs where we also confirm our approach can be orthogonally combined with recent advances in neural ODEs.

Machine-Learning Model for the Prediction of Hypoxaemia during Endoscopic Retrograde Cholangiopancreatography under Monitored Anaesthesia Care.

PURPOSE: Hypoxaemia is a significant adverse event during endoscopic retrograde cholangiopancreatography (ERCP) under monitored anaesthesia care (MAC); however, no model has been developed to predict hypoxaemia. We aimed to develop and compare logistic regression (LR) and machine learning (ML) models to predict hypoxaemia during ERCP under MAC. MATERIALS AND METHODS: We collected patient data from our institutional ERCP database. The study population was randomly divided into training and test sets (7:3). Models were fit to training data and evaluated on unseen test data. The training set was further split into k-fold (k=5) for tuning hyperparameters, such as feature selection and early stopping. Models were trained over k loops; the i-th fold was set aside as a validation set in the i-th loop. Model performance was measured using area under the curve (AUC). RESULTS: We identified 6114 cases of ERCP under MAC, with a total hypoxaemia rate of 5.9%. The LR model was established by combining eight variables and had a test AUC of 0.693. The ML and LR models were evaluated on 30 independent data splits. The average test AUC for LR was 0.7230, which improved to 0.7336 by adding eight more variables with an l1 regularisation-based selection technique and ensembling the LRs and gradient boosting algorithm (GBM). The high-risk group was discriminated using the GBM ensemble model, with a sensitivity and specificity of 63.6% and 72.2%, respectively. CONCLUSION: We established GBM ensemble model and LR model for risk prediction, which demonstrated good potential for preventing hypoxaemia during ERCP under MAC.

A Review of Multivariate Distributions for Count Data Derived from the Poisson Distribution.

The Poisson distribution has been widely studied and used for modeling univariate count-valued data. Multivariate generalizations of the Poisson distribution that permit dependencies, however, have been far less popular. Yet, real-world high-dimensional count-valued data found in word counts, genomics, and crime statistics, for example, exhibit rich dependencies, and motivate the need for multivariate distributions that can appropriately model this data. We review multivariate distributions derived from the univariate Poisson, categorizing these models into three main classes: 1) where the marginal distributions are Poisson, 2) where the joint distribution is a mixture of independent multivariate Poisson distributions, and 3) where the node-conditional distributions are derived from the Poisson. We discuss the development of multiple instances of these classes and compare the models in terms of interpretability and theory. Then, we empirically compare multiple models from each class on three real-world datasets that have varying data characteristics from different domains, namely traffic accident data, biological next generation sequencing data, and text data. These empirical experiments develop intuition about the comparative advantages and disadvantages of each class of multivariate distribution that was derived from the Poisson. Finally, we suggest new research directions as explored in the subsequent discussion section.

XMRF: an R package to fit Markov Networks to high-throughput genetics data.

BACKGROUND: Technological advances in medicine have led to a rapid proliferation of high-throughput "omics" data. Tools to mine this data and discover disrupted disease networks are needed as they hold the key to understanding complicated interactions between genes, mutations and aberrations, and epi-genetic markers. RESULTS: We developed an R software package, XMRF, that can be used to fit Markov Networks to various types of high-throughput genomics data. Encoding the models and estimation techniques of the recently proposed exponential family Markov Random Fields (Yang et al., 2012), our software can be used to learn genetic networks from RNA-sequencing data (counts via Poisson graphical models), mutation and copy number variation data (categorical via Ising models), and methylation data (continuous via Gaussian graphical models). CONCLUSIONS: XMRF is the only tool that allows network structure learning using the native distribution of the data instead of the standard Gaussian. Moreover, the parallelization feature of the implemented algorithms computes the large-scale biological networks efficiently. XMRF is available from CRAN and Github ( https://github.com/zhandong/XMRF ).

Graphical Models via Univariate Exponential Family Distributions.

Undirected graphical models, or Markov networks, are a popular class of statistical models, used in a wide variety of applications. Popular instances of this class include Gaussian graphical models and Ising models. In many settings, however, it might not be clear which subclass of graphical models to use, particularly for non-Gaussian and non-categorical data. In this paper, we consider a general sub-class of graphical models where the node-wise conditional distributions arise from exponential families. This allows us to derive multivariate graphical model distributions from univariate exponential family distributions, such as the Poisson, negative binomial, and exponential distributions. Our key contributions include a class of M-estimators to fit these graphical model distributions; and rigorous statistical analysis showing that these M-estimators recover the true graphical model structure exactly, with high probability. We provide examples of genomic and proteomic networks learned via instances of our class of graphical models derived from Poisson and exponential distributions.

Genomic Alterations in the RB Pathway Indicate Prognostic Outcomes of Early-Stage Lung Adenocarcinoma.

PURPOSE: To better understand the complete genomic architecture of lung adenocarcinoma. EXPERIMENTAL DESIGN: We used array experiments to determine copy number variations and sequenced the complete exomes of the 247 lung adenocarcinoma tumor samples along with matched normal cells obtained from the same patients. Fully annotated clinical data were also available, providing an unprecedented opportunity to assess the impact of genomic alterations on clinical outcomes. RESULTS: We discovered that genomic alternations in the RB pathway are associated with significantly shorter disease-free survival in early-stage lung adenocarcinoma patients. This association was also observed in our independent validation cohort. The current treatment guidelines for early-stage lung adenocarcinoma patients recommend follow-up without adjuvant therapy after complete resection, except for high-risk patients. However, our findings raise the interesting possibility that additional clinical interventions might provide medical benefits to early-stage lung adenocarcinoma patients with genomic alterations in the RB pathway. When examining the association between genomic mutation and histologic subtype, we uncovered the characteristic genomic signatures of various histologic subtypes. Notably, the solid and the micropapillary subtypes demonstrated great diversity in the mutated genes, while the mucinous subtype exhibited the most unique landscape. This suggests that a more tailored therapeutic approach should be used to treat patients with lung adenocarcinoma. CONCLUSIONS: Our analysis of the genomic and clinical data for 247 lung adenocarcinomas should help provide a more comprehensive genomic portrait of lung adenocarcinoma, define molecular signatures of lung adenocarcinoma subtypes, and lead to the discovery of useful prognostic markers that could be used in personalized treatments for early-stage lung adenocarcinoma patients.