Anna Maria Bianucci

Application of Cascade Correlation Networks for Structures toChemistry

We present the application of Cascade Correlation for structures to QSPR (quantitative structure-property relationships) and QSAR (quantitative structure-activity relationships) analysis. Cascade Correlation for structures is a neural network model recently proposed for the processing of structured data. This allows the direct treatment of chemical compounds as labeled trees, which constitutes a novel approach to QSPR/QSAR. We report the results obtained for QSPR on Alkanes (predicting the boiling point) and QSAR of a class of Benzodiazepines. Our approach compares favorably versus the traditional QSAR treatment based on equations and it is competitive with ‘ad hoc’ MLPs for the QSPR problem.

Analysis of the Internal Representations Developed by Neural Networks for Structures Applied to Quantitative Structure-Activity Relationship Studies of Benzodiazepines

An application of recursive cascade correlation (CC) neural networks to quantitative structure-activity relationship (QSAR) studies is presented, with emphasis on the study of the internal representations developed by the neural networks. Recursive CC is a neural network model recently proposed for the processing of structured data. It allows the direct handling of chemical compounds as labeled ordered directed graphs, and constitutes a novel approach to QSAR. The adopted representation of molecular structure captures, in a quite general and flexible way, significant topological aspects and chemical functionalities for each specific class of molecules showing a particular chemical reactivity or biological activity. A class of 1,4-benzodiazepin-2-ones is analyzed by the proposed approach. It compares favorably versus the traditional QSAR treatment based on equations. To show the ability of the model in capturing most of the structural features that account for the biological activity, the internal representations developed by the networks are analyzed by principal component analysis. This analysis shows that the networks are able to discover relevant structural features just on the basis of the association between the molecular morphology and the target property (affinity).

A Novel Approach to QSPR/QSAR Based on Neural Networks for Structures

We present a novel approach based on neural networks for structures to QSPR (quantitative structure-property relationships) and QSAR (quantitative structure-activity relationships) analysis. We face two quite different chemical applications using the same model, i.e. predicting the boiling point of a class of alkanes and QSAR of a class of benzodiazepines. The model, Cascade Correlation for structures, is a class of recursive neural networks recently proposed for the processing of structured domains. Through the use of this model we can represent and process the structure of chemical compounds in the form of labeled trees. We report our results on preliminary applications to show that the model, adopting this representation of molecular structure, can adaptively capture significant topological aspects and chemical fnnctionalities for each specific class of the molecules, just on the basis of the association between the molecular morphology and the target property.

Development of QSAR models for predicting hepatocarcinogenic toxicity of chemicals

A dataset comprising 55 chemicals with hepatocarcinogenic potency indices was collected from the Carcinogenic Potency Database with the aim of developing QSAR models enabling prediction of the above unwanted property for New Chemical Entities. The dataset was rationally split into training and test sets by means of a sphere-exclusion type algorithm. Among the many algorithms explored to search regression models, only a Support Vector Machine (SVM) method led to a QSAR model, which was proved to pass rigorous validation criteria, in accordance with the OECD guidelines. The proposed model is capable to explain the hepatocarcinogenic toxicity and could be exploited for predicting this property for chemicals at the early stage of their development, so optimizing resources and reducing animal testing.

Planarians as a Model to Assess In Vivo the Role of Matrix Metalloproteinase Genes during Homeostasis and Regeneration

Matrix metalloproteinases (MMPs) are major executors of extracellular matrix remodeling and, consequently, play key roles in the response of cells to their microenvironment. The experimentally accessible stem cell population and the robust regenerative capabilities of planarians offer an ideal model to study how modulation of the proteolytic system in the extracellular environment affects cell behavior in vivo. Genome-wide identification of Schmidtea mediterranea MMPs reveals that planarians possess four mmp-like genes. Two of them (mmp1 and mmp2) are strongly expressed in a subset of secretory cells and encode putative matrilysins. The other genes (mt-mmpA and mt-mmpB) are widely expressed in postmitotic cells and appear structurally related to membrane-type MMPs. These genes are conserved in the planarian Dugesia japonica. Here we explore the role of the planarian mmp genes by RNA interference (RNAi) during tissue homeostasis and regeneration. Our analyses identify essential functions for two of them. Following inhibition of mmp1 planarians display dramatic disruption of tissues architecture and significant decrease in cell death. These results suggest that mmp1 controls tissue turnover, modulating survival of postmitotic cells. Unexpectedly, the ability to regenerate is unaffected by mmp1(RNAi). Silencing of mt-mmpA alters tissue integrity and delays blastema growth, without affecting proliferation of stem cells. Our data support the possibility that the activity of this protease modulates cell migration and regulates anoikis, with a consequent pivotal role in tissue homeostasis and regeneration. Our data provide evidence of the involvement of specific MMPs in tissue homeostasis and regeneration and demonstrate that the behavior of planarian stem cells is critically dependent on the microenvironment surrounding these cells. Studying MMPs function in the planarian model provides evidence on how individual proteases work in vivo in adult tissues. These results have high potential to generate significant information for development of regenerative and anti cancer therapies.

Predictive models, based on classification algorithms, for compounds potentially active as mitochondrial ATP-sensitive potassium channel openers

Heart mitochondrial ATP-sensitive potassium channel (mito-K(ATP) channels) are deeply implicated in the self-defense mechanism of ischemic preconditioning. Therefore, exogenous molecules activating these channels are considered as a promising pharmacological tool to reduce the myocardial injury deriving from ischemia/reperfusion events. In our laboratory, a series of 4-spiro-substituted benzopyran derivatives were earlier synthesized and some of them exhibited anti-ischemic properties. In this study, the above compounds are exploited in order to develop QSAR models, based on classification approaches, capable of discriminating between the ones acting as cardioprotective agents and those that are unable to elicit such a property. Molecules belonging to the whole dataset were subjected to CODESSA and E-Dragon calculations in order to compute a large number of molecular descriptors enabling the construction of classification models. Based on the two program packages used, two different experiments were carried out, with the aim of identify batteries of models to be exploited for designing new cardioprotective agents from libraries of new chemical entities. Both model batteries satisfy the rigorous criteria adopted for the validation, either when tested on the training and test set, according to the most straightforward protocol, and when tested on an additional prediction set. They were proven to ensure successful applications in the field of cardioprotective agent design.

Quantitative structure-activity relationships of Benzodiazepines by recursive cascade correlation

An application of recursive cascade correlation to the quantitative structure-activity relationships (QSAR) of a class of Beneodiazepines is presented. Recursive cascade correlation is a neural network model recently proposed for the processing of structured data. This allows the direct treatment of the chemical compounds as labeled ordered trees, which constitutes a novel approach to QSAR. Our approach compares favorably versus the traditional QSAR treatment based on equations.

A 3D Model for the Human Hepatic Asialoglycoprotein Receptor (ASGP-R)

Abstract The human hepatic Asialoglycoprotein Receptor (ASGP-R) consists of two different types of liver specific membrane glycoproteins that bind to terminal galactose and acetylgalactosamine residues of serum glycoproteins. The two different polypeptide chains are referred to as two receptor subunits, HH1 and HH2, which are both involved in the activity of the functional receptor. This receptor has served as a model for understanding receptor-mediated endocytosis and carbohydrate mediated recognition phenomena. Here models for the C-terminal extracellular region of both HH1 and HH2 subunit are presented. The standard homology building procedure was modified in order to make it suitable for the modeling problem at hand. The models for the extracellular regions of HH1 and HH2 were initially constructed by exploiting several fragments, belonging to proteins of known 3D structure, and showing high local sequence similarity with respect to the glycoproteins of interest. Putative binding sites were first hypothesized on the basis of the comparison with other complexes of lectins, the crystal structure of which was available in the Protein Data Bank. A model for the complex involving the HH2 subunit and the typical high affinity ligand N-acetylgalactosamine (NacGal) was refined as the first by a suitable combination of MD simulations and Energy Minimization calculations, since it seemed to quickly converge to a plausible structure. An intermediate model for HH1 was then rebuilt on the basis of the refined model for HH2. It was then submitted to a sequence of molecular dynamics simulations with templates which took into account the secondary structure prediction for a final refinement. The structures of small regions of the models, located around the binding sites, were compared with more recent crystallographic data regarding a complex involving the mutant of Mannose Binding Protein QPDWGH (1BCH entry in the Protein Data Bank) and NacGal. This mutant shows high local sequence similarity with HH1 and HH2 at the binding sites. On the basis of the above comparison, different locations of the binding sites were also considered. In addition to other expected interactions, two hydrophobic interactions were observed in the models with Trp residues (positions 243 in HH1 and 181 or 267 in HH2 respectively) and His residues (positions 256 in HH1 and 184 in HH2 respectively). The quality of the models was evaluated by the Procheck program and they seemed plausible. This observation together with analogies found between binding sites of the models and 1BCH supported the validity of the models. A further validation element arose by comparison between experimental binding data available in the literature about the homologous rat receptor subunits and theoretical interaction energies evaluated, by means of the DOCK 3.5 program, in models for the rat subunits obtained from the corresponding human ones. The new modeling procedure used here appears to be a well-suited method for structural analysis of small regions, located around the ligands, in proteins of unknown 3D structure.

Synthesis and biological activities of vitamin D-like inhibitors of CYP24 hydroxylase

Selective inhibitors of CYP24A1 represent an important synthetic target in a search for novel vitamin D compounds of therapeutic value. In the present work, we show the synthesis and biological properties of two novel side chain modified 2-methylene-19-nor-1,25(OH)(2)D(3) analogs, the 22-imidazole-1-yl derivative 2 (VIMI) and the 25-N-cyclopropylamine compound 3 (CPA1), which were efficiently prepared in convergent syntheses utilizing the Lythgoe type Horner-Wittig olefination reaction. When tested in a cell-free assay, both compounds were found to be potent competitive inhibitors of CYP24A1, with the cyclopropylamine analog 3 exhibiting an 80-1 selective inhibition of CYP24A1 over CYP27B1. Addition of 3 to a mouse osteoblast culture sustained the level of 1,25(OH)(2)D(3), further demonstrating its effectiveness in CYP24A1 inhibition. Importantly, the in vitro effects on human promyeloid leukemia (HL-60) cell differentiation by 3 were nearly identical to those of 1,25(OH)(2)D(3) and in vivo the compound showed low calcemic activity. Finally, the results of preliminary theoretical studies provide useful insights to rationalize the ability of analog 3 to selectively inhibit the cytochrome P450 isoform CYP24A1.

QSAR studies on BK channel activators

QSAR studies were developed on the basis of a dataset comprising BK channel activators previously synthesized and biologically assayed in our laboratory, in order to obtain highly accurate models enabling prediction of affinity toward the channel for New Chemical Entities (NCEs). Many molecular descriptors were computed by the CODESSA software. They were initially exploited in order to rationally split the available dataset into training and test set pairs, which supplied the basis for the development of QSAR models. Models were subjected to rigorous validation analysis based on the estimate of several statistical parameters, for the seek of the most accurate and simplest model enabling prediction of BK channel affinity.