Classification of Breast Cancer Histopathological Images Using DenseNet and Transfer Learning.

Breast cancer is one of the most common invading cancers in women. Analyzing breast cancer is nontrivial and may lead to disagreements among experts. Although deep learning methods achieved an excellent performance in classification tasks including breast cancer histopathological images, the existing state-of-the-art methods are computationally expensive and may overfit due to extracting features from in-distribution images. In this paper, our contribution is mainly twofold. First, we perform a short survey on deep-learning-based models for classifying histopathological images to investigate the most popular and optimized training-testing ratios. Our findings reveal that the most popular training-testing ratio for histopathological image classification is 70%: 30%, whereas the best performance (e.g., accuracy) is achieved by using the training-testing ratio of 80%: 20% on an identical dataset. Second, we propose a method named DenTnet to classify breast cancer histopathological images chiefly. DenTnet utilizes the principle of transfer learning to solve the problem of extracting features from the same distribution using DenseNet as a backbone model. The proposed DenTnet method is shown to be superior in comparison to a number of leading deep learning methods in terms of detection accuracy (up to 99.28% on BreaKHis dataset deeming training-testing ratio of 80%: 20%) with good generalization ability and computational speed. The limitation of existing methods including the requirement of high computation and utilization of the same feature distribution is mitigated by dint of the DenTnet.

Concentration effects on the self-assembly of tyrosine molecules.

Molecular self-assembly is a ubiquitous phenomenon in which individual atoms or molecules set up an ordered structure. It is of high interest for understanding the biology and a variety of diseases at the molecular level. In this work, we studied the self-assembly of tyrosine molecules via extensive molecular dynamics simulations. The formation of structures by self-assembly was systematically studied at various concentrations, from very low to very high. The temperature was kept constant, at which, in our former studies, we have already observed well-formed self-assembled structures. Depending on the concentration, the system displays a wide range of different structures, ranging from freely scattered monomers to very well formed four-fold structures. Different regimes of concentration dependence are observed. The results are proved by calculating the moments of inertia of the structures and the number of hydrogen bonds formed. Free energy landscapes calculated for the number of hydrogen bonds versus the number of contacts within a criterion provide insights into the structures observed.

Self-assembly of aromatic amino acids: a molecular dynamics study.

The self assembly processes of aromatic amino acids, phenylalanine, tyrosine, and tryptophan have been simulated and were observed to form fibril-like aggregates linked to certain rare diseases and instances of biological membrane disruption. Pure systems and their mixtures were studied systematically at constant temperatures and free energy landscapes were produced describing the height and the number of assembled monomers associated with lower energy structures. Consistent with some previous work, aromatic amino acid monomers display a tendency to arrange with a four-fold symmetry. The occurrence of this and other ordered structures increases at higher temperatures. At lower temperatures our binary mixture simulations indicate that increasing tryptophan content drives the assembly process away from the formation of distinct nanostructures and toward disordered aggregates which is in line with experimental observations of pure tryptophan solutions. This work provides molecular level insight to a variety of different physical phenomena relevant to fields including human disease.

Self-assembly of phenylalanine-based molecules.

Using molecular dynamics, we study the self-assembly of phenylalanine with charged end-groups at various temperatures and concentrations. As in the case of diphenylalanine, we observe the formation of nanotubes; however, phenylalanine aggregates in layers of four, not six, molecules. The observed aggregates are consistent with recent experimental measurements of fibrils obtained from mice with phenylketonuria. We investigate the stability and the mechanism by which these tubular structures form and discuss potential toxicity mechanisms.

Conformational transitions of a weak polyampholyte.

Using grand canonical Monte Carlo simulations of a flexible polyelectrolyte where the charges are in contact with a reservoir of constant chemical potential given by the solution pH, we study the behavior of weak polyelectrolytes in poor and good solvent conditions for polymer backbone. We address the titration behavior and conformational properties of a flexible diblock polyampholyte chain formed of two oppositely charged weak polyelectrolyte blocks, each containing equal number of identical monomers. The change of solution pH induces charge asymmetry in a diblock polyampholyte. For diblock polyampholyte chains in poor solvents, we demonstrate that a discontinuous transition between extended (tadpole) and collapsed (globular) conformational states is attainable by varying the solution pH. The double-minima structure in the probability distribution of the free energy provides direct evidence for the first-order like nature of this transition. At the isoelectric point electrostatically driven coil-globule transition of diblock polyampholytes in good solvents is found to consist of different regimes identified with increasing electrostatic interaction strength. At pH values above or below the isoelectric point diblock chains are found to have polyelectrolyte-like behavior due to repulsion between uncompensated charges along the chain.

Multicanonical Monte Carlo simulations of a de novo designed protein with end-to-end ß-sheet.

One of the smallest proteins with end-to-end ß-sheet is the designed 36-residue protein DS119. We recently suggested that the rate-limiting step in the folding of the ßαß protein is the formation of the central helix that then provides a scaffold for the parallel ß-sheet formed by the two chain ends. In the present report we investigate whether and how this folding mechanism depends on the energy function, and compare the efficiency of molecular dynamics and Monte Carlo implementations of multicanonical sampling. While we find the native structure with similar frequency as in our previous simulations, we observe that the folding mechanism differs for both force fields.