Recent advances in pharmacogenomics have generated a wealth of data of different types whose analysis have helped in the identification of signatures of different cellular sensitivity/resistance responses to hundreds of chemical compounds. Among the different data types, gene expression has proven to be the more successful for the inference of drug response in cancer cell lines. Although effective, the whole transcriptome can introduce noise in the predictive models, since specific mechanisms are required for different drugs and these realistically involve only part of the proteins encoded in the genome. We analyzed the pharmacogenomics data of 961 cell lines tested with 265 anti-cancer drugs and developed different machine learning approaches for dissecting the genome systematically and predict drug responses using both drug-unspecific and drug-specific genes. These methodologies reach better response predictions for the vast majority of the screened drugs using tens to few hundreds genes specific to each drug instead of the whole genome, thus allowing a better understanding and interpretation of drug-specific response mechanisms which are not necessarily restricted to the drug known targets.

ObjectiveThe increased survival in Systemic Lupus Erythematosus (SLE) patients implies the development of chronic damage, occurring in up to 50% of cases. Its prevention is a major goal in the SLE management. We aimed at predicting chronic damage in a large monocentric SLE cohort by using neural networks.MethodsWe enrolled 413 SLE patients (M/F 30/383; mean age ± SD 46.3±11.9 years; mean disease duration ± SD 174.6 ± 112.1 months). Chronic damage was assessed by the SLICC/ACR Damage Index (SDI). We applied Recurrent Neural Networks (RNNs) as a machine-learning model to predict the risk of chronic damage. The clinical data sequences registered for each patient during the follow-up were used for building and testing the RNNs.ResultsAt the first visit in the Lupus Clinic, 35.8% of patients had an SDI≥1. For the RNN model, two groups of patients were analyzed: patients with SDI = 0 at the baseline, developing damage during the follow-up (N = 38), and patients without damage (SDI = 0). We created a mathematical model with an AUC value of 0.77, able to predict damage development. A threshold value of 0.35 (sensitivity 0.74, specificity 0.76) seemed able to identify patients at risk to develop damage.ConclusionWe applied RNNs to identify a prediction model for SLE chronic damage. The use of the longitudinal data from the Sapienza Lupus Cohort, including laboratory and clinical items, resulted able to construct a mathematical model, potentially identifying patients at risk to develop damage.

ObjectiveLimited evidences are available on biomarkers to recognize Systemic Lupus erythematosus (SLE) patients at risk to develop erosive arthritis. Anti-citrullinated peptide antibodies (ACPA) have been widely investigated and identified in up to 50% of X-ray detected erosive arthritis; conversely, few studies evaluated anti-carbamylated proteins antibodies (anti-CarP). Here, we considered the application of machine learning models to identify relevant factors in the development of ultrasonography (US)-detected erosive damage in a large cohort of SLE patients with joint involvement.MethodsWe enrolled consecutive SLE patients with arthritis/arthralgia. All patients underwent joint (DAS28, STR) and laboratory assessment (detection of ACPA, anti-CarP, Rheumatoid Factor, SLE-related antibodies). The bone surfaces of metacarpophalangeal and proximal interphalangeal joints were assessed by US: the presence of erosions was registered with a dichotomous value (0/1), obtaining a total score (0–20). Concerning machine learning techniques, we applied and compared Logistic Regression and Decision Trees in conjunction with the feature selection Forward Wrapper method.ResultsWe enrolled 120 SLE patients [M/F 8/112, median age 47.0 years (IQR 15.0); median disease duration 120.0 months (IQR 156.0)], 73.3% of them referring at least one episode of arthritis. Erosive damage was identified in 25.8% of patients (mean±SD 0.7±1.6), all of them with clinically evident arthritis. We applied Logistic Regression in conjunction with the Forward Wrapper method, obtaining an AUC value of 0.806±0.02. As a result of the learning procedure, we evaluated the relevance of the different factors: this value was higher than 35% for ACPA and anti-CarP.ConclusionThe application of Machine Learning Models allowed to identify factors associated with US-detected erosive bone damage in a large SLE cohort and their relevance in determining this phenotype. Although the scope of this study is limited by the small sample size and its cross-sectional nature, the results suggest the relevance of ACPA and anti-CarP antibodies in the development of erosive damage as also pointed out in other studies.

In the present paper we propose to rewrite a nonsmooth problem subjected to convex constraints as an unconstrained problem. We show that this novel formulation shares the same global and local minima with the original constrained problem. Moreover, the reformulation can be solved with standard nonsmooth optimization methods if we are able to make projections onto the feasible sets. Numerical evidence shows that the proposed formulation compares favorably against state-of-art approaches. Code can be found at https://github.com/jth3galv/dfppm.

In liberalized markets, there usually exists a day-ahead session where energy is sold and acquired for the following production day. Owing to the high uncertainty of its production, renewable energy (wind in particular) can significantly influence the network imbalance of the following day.In this work, we consider the problem of predicting the sum of the bid volumes for wind energy of all the producers inside the day-ahead energy market. This is a valuable tool to be used by an energy provider in order to determine the imbalance of a market zone and, thus, properly size its bids. In particular, we focus on the estimation of the possible relationship between the meteorological forecasts and the wind power offered on the market by the companies for a market zone. We propose a machine learning model which is used to compute a 1-day-ahead forecast.The input-output mapping is obtained by support vector regression. The input feature vector is defined by a suitable feature extraction technique since the meteorological forecasts are given on a lattice of thousands of geographical points. The computational experiments are performed considering the Italian market as a case study (years 2012-2016). The results show that the proposed feature extraction technique, selecting only some geographical zones, manages to reduce the error attained using all the features. Moreover, classical statistical methods are shown to be outperformed by machine learning models. The analysis reveals also some weaknesses of the model, which may be due to other nonmeteorological factors at play. KEYWORDS day-ahead forecasting, extreme learning machines (ELMs), feature selection, machine learning, support vector regression (SVR), wind power market INTRODUCTIONOver the last few years, there has been an increasing interest in the study of numerical predictive models related to renewable energies, which are a serious alternative to replace traditional energy sources. Renewable energies represent a suitable and economical alternative to conventional power plants, but they are characterized by a high degree of variability. To deal with these uncertainties, several machine learning and feature selection techniques have been adopted for various forecasting tasks related to renewable energy.One of the most important and widely studied problems is the forecasting of energy production of power plants. Machine learning techniques have been widely used to obtain a mapping from weather data, generally wind speed, to the power generated by the turbines of a power plant. In the work of Heinermann and Kramer, 1 2 machine learning models, namely, regression trees and support vector regression (SVR), are combined with an ensemble method. Support vector regression is used also in the work of Zeng and Qiao, 2 where a new kernel, based on the wavelet mother function, is shown to outperform the more commonly used radial basis function kernel. In the work of Wan et al, 3 extreme learning machines (ELMs) are used as a fast and reliable alternative to the more traditional ...