Alireza Sadeghian

Ashirbani Saha, Pantea Fadaiefard, Jessica E. Rabski et al.

We performed a PubMed search to find 148 papers published between January 2010 and December 2019 related to human brain, Diffusion Tensor Imaging (DTI), and Machine Learning (ML). The studies focused on healthy cohorts (n = 15), mental health disorders (n = 25), tumor (n = 19), trauma (n = 5), dementia (n = 24), developmental disorders (n = 5), movement disorders (n = 9), other neurological disorders (n = 27), miscellaneous non-neurological disorders, or without stating the disease of focus (n = 7), and multiple combinations of the aforementioned categories (n = 12). Classification of patients using information from DTI stands out to be the most commonly (n = 114) performed ML application. A significant number (n = 93) of studies used support vector machines (SVM) as the preferred choice of ML model for classification. A significant portion (31/44) of publications in the recent years (2018-2019) continued to use SVM, support vector regression, and random forest which are a part of traditional ML. Though many types of applications across various health conditions (including healthy) were conducted, majority of the studies were based on small cohorts (less than 100) and did not conduct independent/external validation on test sets.

Lorenzo Livi, Alessandro Giuliani, Alireza Sadeghian

This paper deals with the relations among structural, topological, and chemical properties of the E.Coli proteome from the vantage point of the solubility/aggregation propensity of proteins. Each E.Coli protein is initially represented according to its known folded 3D shape. This step consists in representing the available E.Coli proteins in terms of graphs. We first analyze those graphs by considering pure topological characterizations, i.e., by analyzing the mass fractal dimension and the distribution underlying both shortest paths and vertex degrees. Results confirm the general architectural principles of proteins. Successively, we focus on the statistical properties of a representation of such graphs in terms of vectors composed of several numerical features, which we extracted from their structural representation. We found that protein size is the main discriminator for the solubility, while however there are other factors that help explaining the solubility degree. We finally analyze such data through a novel one-class classifier, with the aim of discriminating among very and poorly soluble proteins. Results are encouraging and consolidate the potential of pattern recognition techniques when employed to describe complex biological systems.