J. P. Doucet

Bio: J. P. Doucet is an academic researcher from Paris Diderot University. The author has contributed to research in topics: Support vector machine & Projection pursuit regression. The author has an hindex of 6, co-authored 7 publications receiving 99 citations.

Investigation of the influence of protein corona composition on gold nanoparticle bioactivity using machine learning approaches.

Abstract: The understanding of the mechanisms and interactions that occur when nanomaterials enter biological systems is important to improve their future use. The adsorption of proteins from biological fluids in a physiological environment to form a corona on the surface of nanoparticles represents a key step that influences nanoparticle behaviour. In this study, the quantitative description of the composition of the protein corona was used to study the effect on cell association induced by 84 surface-modified gold nanoparticles of different sizes. Quantitative relationships between the protein corona and the activity of the gold nanoparticles were modelled by using several machine learning-based linear and non-linear approaches. Models based on a selection of only six serum proteins had robust and predictive results. The Projection Pursuit Regression method had the best performances (r(2) = 0.91; Q(2)loo = 0.81; r(2)ext = 0.79). The present study confirmed the utility of protein corona composition to predict the bioactivity of gold nanoparticles and identified the main proteins that act as promoters or inhibitors of cell association. In addition, the comparison of several techniques showed which strategies offer the best results in prediction and could be used to support new toxicological studies on gold-based nanomaterials.

Linear and non-linear modelling of the cytotoxicity of TiO2 and ZnO nanoparticles by empirical descriptors

Abstract: Titanium oxide (TiO2) and zinc oxide (ZnO) nanoparticles are among the most widely used in different applications in daily life. In this study, local regression and classification models were developed for a set of ZnO and TiO2 nanoparticles tested at different concentrations for their ability to disrupt the lipid membrane in cells. Different regression techniques were applied and compared by checking the robustness of the models and their external predictive ability. Additionally, a simple classification model was developed, which predicts the potential for disruption of the studied nanoparticles with good accuracy (overall accuracy, specificity, and sensitivity >80%) on the basis of two empirical descriptors. The present study demonstrates that empirical descriptors, such as experimentally determined size and tested concentrations, are relevant to modelling the activity of nanoparticles. This information may be useful to screen the potential for harmful effect of nanoparticles in different experimental conditions and to optimize the design of toxicological tests. Results from the present study are useful to support and refine the future application of in silico tools to nanoparticles, for research and regulatory purposes.

QSAR models for predicting the toxicity of piperidine derivatives against Aedes aegypti.

Abstract: QSAR models are proposed for predicting the toxicity of 33 piperidine derivatives against Aedes aegypti. From 2D topological descriptors, calculated with the PaDEL software, ordinary least squares multilinear regression (OLS-MLR) treatment from the QSARINS software and machine learning and related approaches including linear and radial support vector machine (SVM), projection pursuit regression (PPR), radial basis function neural network (RBFNN), general regression neural network (GRNN) and k-nearest neighbours (k-NN), led to four-variable models. Their robustness and predictive ability were evaluated through both internal and external validation. Determination coefficients (r2) greater than 0.85 on the training sets and 0.8 on the test sets were obtained with OLS-MLR and linear SVM. They slightly outperform PPR, radial SVM and RBFNN, whereas GRNN and k-NN showed lower performance. The easy availability of the involved structural descriptors and the simplicity of the MLR model make the corresponding model attractive at an exploratory level for proposing, from this limited dataset, guidelines in the design of new potentially active molecules.

Computational approaches for the prediction of the selective uptake of magnetofluorescent nanoparticles into human cells

Abstract: The use of functionalized nanomaterials is of high importance in biomedical applications like the efficient targeting of cancer cells. This paper proposes a comparison of different statistical and mechanistic aspects of new QSAR models generated to predict the selective uptake of a library of surface modified nanoparticles tested in different human cell types. Additionally, a new approach based on the combination of multivariate factorial analysis and QSAR is proposed to generate a 2-dimensional map of the selective uptake of the surface modified nanoparticles into multiple cell types. This map offers an immediate view of the uptake of the nanoparticles, distinguishing among those with high or low uptake in one or more of the studied cells. Finally, QSAR models are generated to predict the coordinates of the studied nanoparticles in the 2D map from their molecular structure. This predictive map is useful to screen new and existing surface modified nanoparticles for diagnostic and biomedical uses.

Structure–activity relationship (SAR) modelling of mosquito larvicides

Abstract: An attempt was made to derive structure–activity models allowing the prediction of the larvicidal activity of structurally diverse chemicals against mosquitoes. A database of 188 chemicals with their activity on Aedes aegypti larvae was constituted from analysis of original publications. The activity values were expressed in log 1/IC50 (concentration required to produce 50% inhibition of larval development, mmol). All the chemicals were encoded by means of CODESSA and autocorrelation descriptors. Partial least squares analysis, classification and regression tree, random forest and boosting regression tree analyses, Kohonen self-organizing maps, linear artificial neural networks, three-layer perceptrons, radial basis function artificial neural networks and support vector machines with linear, polynomial, radial basis function and sigmoid kernels were tested as statistical tools. Because quantitative models did not give good results, a two-class model was designed. The three-layer perceptron significantly o...

Machine learning

Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning provides predictive analysis into silver nanoparticle protein corona formation from physicochemical properties

Abstract: Proteins encountered in biological and environmental systems bind to engineered nanomaterials (ENMs) to form a protein corona (PC) that alters the surface chemistry, reactivity, and fate of the ENMs. Complexities such as the diversity of the PC and variation with ENM properties and reaction conditions make the PC population difficult to predict. Here, we support the development of predictive models for PC populations by relating biophysicochemical characteristics of proteins, ENMs, and solution conditions to PC formation using random forest classification. The resulting model offers a predictive analysis into the population of PC proteins in Ag ENM systems of various ENM size and surface coatings. With an area under the receiver operating characteristic curve of 0.83 and F1-score of 0.81, a model with strong performance has been constructed based upon experimental data. The weighted contribution of each variable provides recommendations for mechanistic models based upon protein enrichment classification results. Protein biophysical properties such as pI and weight are weighted heavily. Yet, ENM size, surface charge, and solution ionic strength also proved essential to an accurate model. The model can be readily modified and applied to other ENM PC populations. The model presented here represents the first step toward robust predictions of PC fingerprints.