Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

A Multivariate Poisson Deep Learning Model for Genomic Prediction of Count Data.

In G3 (Bethesda, Md.)

The paradigm called genomic selection (GS) is a revolutionary way of developing new plants and animals. This is a predictive methodology, since it uses learning methods to perform its task. Unfortunately, there is no universal model that can be used for all types of predictions; for this reason, specific methodologies are required for each type of output (response variables). Since there is a lack of efficient methodologies for multivariate count data outcomes, in this paper, a multivariate Poisson deep neural network (MPDN) model is proposed for the genomic prediction of various count outcomes simultaneously. The MPDN model uses the minus log-likelihood of a Poisson distribution as a loss function, in hidden layers for capturing nonlinear patterns using the rectified linear unit (RELU) activation function and, in the output layer, the exponential activation function was used for producing outputs on the same scale of counts. The proposed MPDN model was compared to conventional generalized Poisson regression models and univariate Poisson deep learning models in two experimental data sets of count data. We found that the proposed MPDL outperformed univariate Poisson deep neural network models, but did not outperform, in terms of prediction, the univariate generalized Poisson regression models. All deep learning models were implemented in Tensorflow as back-end and Keras as front-end, which allows implementing these models on moderate and large data sets, which is a significant advantage over previous GS models for multivariate count data.

Montesinos-López Osval Antonio, Montesinos-López José Cricelio, Singh Pawan, Lozano-Ramirez Nerida, Barrón-López Alberto, Montesinos-López Abelardo, Crossa José


Genomic selection and genomic prediction, Poisson regression, count data of wheat lines, multivariate Poisson deep neural network, univariate Poisson deep neural network

General General

Machine learning predicts stem cell transplant response in severe scleroderma.

In Annals of the rheumatic diseases ; h5-index 121.0

OBJECTIVE : The Scleroderma: Cyclophosphamide or Transplantation (SCOT) trial demonstrated clinical benefit of haematopoietic stem cell transplant (HSCT) compared with cyclophosphamide (CYC). We mapped PBC (peripheral blood cell) samples from the SCOT clinical trial to scleroderma intrinsic subsets and tested the hypothesis that they predict long-term response to HSCT.

METHODS : We analysed gene expression from PBCs of SCOT participants to identify differential treatment response. PBC gene expression data were generated from 63 SCOT participants at baseline and follow-up timepoints. Participants who completed treatment protocol were stratified by intrinsic gene expression subsets at baseline, evaluated for event-free survival (EFS) and analysed for differentially expressed genes (DEGs).

RESULTS : Participants from the fibroproliferative subset on HSCT experienced significant improvement in EFS compared with fibroproliferative participants on CYC (p=0.0091). In contrast, EFS did not significantly differ between CYC and HSCT arms for the participants from the normal-like subset (p=0.77) or the inflammatory subset (p=0.1). At each timepoint, we observed considerably more DEGs in HSCT arm compared with CYC arm with HSCT arm showing significant changes in immune response pathways.

CONCLUSIONS : Participants from the fibroproliferative subset showed the most significant long-term benefit from HSCT compared with CYC. This study suggests that intrinsic subset stratification of patients may be used to identify patients with SSc who receive significant benefit from HSCT.

Franks Jennifer M, Martyanov Viktor, Wang Yue, Wood Tammara A, Pinckney Ashley, Crofford Leslie J, Keyes-Elstein Lynette, Furst Daniel E, Goldmuntz Ellen, Mayes Maureen D, McSweeney Peter, Nash Richard A, Sullivan Keith M, Whitfield Michael L


cyclophosphamide, systemic sclerosis, treatment

oncology Oncology

Big Data Solutions for Controversies in Breast Cancer Treatment.

In Clinical breast cancer

The digital world of data is expanding with an annual growth rate of 40%, and health care is among the fastest growing sector of the digital world with an annual growth rate of 48%. Rapid growth in technology has augmented data generation; for example, electronic health records produce huge amounts of patient-level data, whereas national registries capture information on numerous factors affecting health care delivery and patient outcomes. This big data can be utilized to improve health care outcomes. This review discusses relevant applications in breast cancer treatment.

Cobb Adrienne N, Janjua Haroon M, Kuo Paul C


Breast oncology, Genomics, Health disparities, Machine learning, Predictive analytics

Cardiology Cardiology

Noninvasive estimation of aortic hemodynamics and cardiac contractility using machine learning.

In Scientific reports ; h5-index 158.0

Cardiac and aortic characteristics are crucial for cardiovascular disease detection. However, noninvasive estimation of aortic hemodynamics and cardiac contractility is still challenging. This paper investigated the potential of estimating aortic systolic pressure (aSBP), cardiac output (CO), and end-systolic elastance (Ees) from cuff-pressure and pulse wave velocity (PWV) using regression analysis. The importance of incorporating ejection fraction (EF) as additional input for estimating Ees was also assessed. The models, including Random Forest, Support Vector Regressor, Ridge, Gradient Boosting, were trained/validated using synthetic data (n = 4,018) from an in-silico model. When cuff-pressure and PWV were used as inputs, the normalized-RMSEs/correlations for aSBP, CO, and Ees (best-performing models) were 3.36 ± 0.74%/0.99, 7.60 ± 0.68%/0.96, and 16.96 ± 0.64%/0.37, respectively. Using EF as additional input for estimating Ees significantly improved the predictions (7.00 ± 0.78%/0.92). Results showed that the use of noninvasive pressure measurements allows estimating aSBP and CO with acceptable accuracy. In contrast, Ees cannot be predicted from pressure signals alone. Addition of the EF information greatly improves the estimated Ees. Accuracy of the model-derived aSBP compared to in-vivo aSBP (n = 783) was very satisfactory (5.26 ± 2.30%/0.97). Future in-vivo evaluation of CO and Ees estimations remains to be conducted. This novel methodology has potential to improve the noninvasive monitoring of aortic hemodynamics and cardiac contractility.

Bikia Vasiliki, Papaioannou Theodore G, Pagoulatou Stamatia, Rovas Georgios, Oikonomou Evangelos, Siasos Gerasimos, Tousoulis Dimitris, Stergiopulos Nikolaos


General General

A mask R-CNN model for reidentifying extratropical cyclones based on quasi-supervised thought.

In Scientific reports ; h5-index 158.0

The applications of machine learning/deep learning (ML/DL) methods in meteorology have developed considerably in recent years. Massive amounts of meteorological data are conducive to improving the training effect and model performance of ML/DL, but the establishment of training datasets is often time consuming, especially in the context of supervised learning. In this paper, to identify the two-dimensional (2D) structures of extratropical cyclones in the Northern Hemisphere, a quasi-supervised reidentification method for extratropical cyclones is proposed. This method first uses a traditional automatic cyclone identification method to construct a trainable labeled dataset and then reidentifies extratropical cyclones in a quasi-supervised fashion by using a (pre-trained) Mask region-based convolutional neural network (Mask R-CNN) model. In comparison, the new method increases the number of identified cyclones by 8.29%, effectively supplementing the traditional method. The newly recognized cyclones are mainly shallow or moderately deep subsynoptic-scale cyclones. However, a considerable portion of the new cyclones along the coastlines of the oceans are accompanied by strong winds. In addition, the Mask R-CNN model also shows good performance in identifying the horizontal structures of tropical cyclones. The quasi-supervised concept proposed in this paper may shed some light on accurate target identification in other research fields.

Lu Chuhan, Kong Yang, Guan Zhaoyong


General General

Deep learning enabled smart mats as a scalable floor monitoring system.

In Nature communications ; h5-index 260.0

Toward smart building and smart home, floor as one of our most frequently interactive interfaces can be implemented with embedded sensors to extract abundant sensory information without the video-taken concerns. Yet the previously developed floor sensors are normally of small scale, high implementation cost, large power consumption, and complicated device configuration. Here we show a smart floor monitoring system through the integration of self-powered triboelectric floor mats and deep learning-based data analytics. The floor mats are fabricated with unique "identity" electrode patterns using a low-cost and highly scalable screen printing technique, enabling a parallel connection to reduce the system complexity and the deep-learning computational cost. The stepping position, activity status, and identity information can be determined according to the instant sensory data analytics. This developed smart floor technology can establish the foundation using floor as the functional interface for diverse applications in smart building/home, e.g., intelligent automation, healthcare, and security.

Shi Qiongfeng, Zhang Zixuan, He Tianyiyi, Sun Zhongda, Wang Bingjie, Feng Yuqin, Shan Xuechuan, Salam Budiman, Lee Chengkuo