DeepSIBA: chemical structure-based inference of biological alterations using deep learning.

Predicting whether a chemical structure leads to a desired or adverse biological effect can have a significant impact for in silico drug discovery. In this study, we developed a deep learning model where compound structures are represented as graphs and then linked to their biological footprint. To make this complex problem computationally tractable, compound differences were mapped to biological effect alterations using Siamese Graph Convolutional Neural Networks. The proposed model was able to encode molecular graph pairs and identify structurally dissimilar compounds that affect similar biological processes with high precision. Additionally, by utilizing deep ensembles to estimate uncertainty, we were able to provide reliable and accurate predictions for chemical structures that are very different from the ones used during training. Finally, we present a novel inference approach, where the trained models are used to estimate the signaling pathway signature of a compound perturbation, using only its chemical structure as input, and subsequently identify which substructures influenced the predicted pathways. As a use case, this approach was used to infer important substructures and affected signaling pathways of FDA-approved anticancer drugs.

A device for high-throughput monitoring of degradation in soft tissue samples.

This work describes the design and validation of a novel device, the High-Throughput Degradation Monitoring Device (HDD), for monitoring the degradation of 24 soft tissue samples over incubation periods of several days inside a cell culture incubator. The device quantifies sample degradation by monitoring its deformation induced by a static gravity load. Initial instrument design and experimental protocol development focused on quantifying cartilage degeneration. Characterization of measurement errors, caused mainly by thermal transients and by translating the instrument sensor, demonstrated that HDD can quantify sample degradation with <6 µm precision and <10 µm temperature-induced errors. HDD capabilities were evaluated in a pilot study that monitored the degradation of fresh ex vivo human cartilage samples by collagenase solutions over three days. HDD could robustly resolve the effects of collagenase concentration as small as 0.5 mg/ml. Careful sample preparation resulted in measurements that did not suffer from donor-to-donor variation (coefficient of variance <70%). Due to its unique combination of sample throughput, measurement precision, temporal sampling and experimental versality, HDD provides a novel biomechanics-based experimental platform for quantifying the effects of proteins (cytokines, growth factors, enzymes, antibodies) or small molecules on the degradation of soft tissues or tissue engineering constructs. Thereby, HDD can complement established tools and in vitro models in important applications including drug screening and biomaterial development.

Quantifying Cartilage Biomechanical Properties Using a Linearized Frequency-Domain Method.

Articular cartilage function relies on its unique mechanical behavior. Cartilage mechanics have been described by several analytic models, whose parameters are usually estimated by fitting their constitutive equations to stress-relaxation data. This procedure can be long and is prone to experimental and fitting errors. Τhis study describes a novel methodology for estimating the biomechanical properties of cartilage samples based on their linearized frequency response, derived by applying a series of small-amplitude harmonic displacements superimposed to a bias strain. The proposed methodology, denoted as linearized frequency-domain method (LFM), was demonstrated by quantifying the effects of collagenase and hyaluronidase on cartilage, where it provided robust cartilage parameter estimates that overall agreed well with estimates obtained by stress-relaxation analysis. LFM was also applied to unveil the strain-dependent nature of porcine cartilage biomechanical parameters. Results showed that increasing the bias strain from 5% to 15% caused a significant decrease in cartilage permeability but did not have significant effect on the compression modulus and the Poisson's ratio. Apart from cartilage, LFM can potentially quantify the strain-dependent nature of tissues and biomaterials, thereby enhance tissue-level understanding on organ physiology and pathology, lead to better computational tissue models, and guide tissue engineering research.

Network-Based Analysis of Nutraceuticals in Human Hepatocellular Carcinomas Reveals Mechanisms of Chemopreventive Action.

Chronic inflammation is associated with the development of human hepatocellular carcinoma (HCC), an essentially incurable cancer. Anti-inflammatory nutraceuticals have emerged as promising candidates against HCC, yet the mechanisms through which they influence the cell signaling machinery to impose phenotypic changes remain unresolved. Herein we implemented a systems biology approach in HCC cells, based on the integration of cytokine release and phospoproteomic data from high-throughput xMAP Luminex assays to elucidate the action mode of prominent nutraceuticals in terms of topology alterations of HCC-specific signaling networks. An optimization algorithm based on SigNetTrainer, an Integer Linear Programming formulation, was applied to construct networks linking signal transduction to cytokine secretion by combining prior knowledge of protein connectivity with proteomic data. Our analysis identified the most probable target phosphoproteins of interrogated compounds and predicted translational control as a new mechanism underlying their anticytokine action. Induced alterations corroborated with inhibition of HCC-driven angiogenesis and metastasis.

Modeling of signaling pathways in chondrocytes based on phosphoproteomic and cytokine release data.

OBJECTIVE: Chondrocyte signaling is widely identified as a key component in cartilage homeostasis. Dysregulations of the signaling processes in chondrocytes often result in degenerative diseases of the tissue. Traditionally, the literature has focused on the study of major players in chondrocyte signaling, but without considering the cross-talks between them. In this paper, we systematically interrogate the signal transduction pathways in chondrocytes, on both the phosphoproteomic and cytokine release levels. METHODS: The signaling pathways downstream 78 receptors of interest are interrogated. On the phosphoproteomic level, 17 key phosphoproteins are measured upon stimulation with single treatments of 78 ligands. On the cytokine release level, 55 cytokines are measured in the supernatant upon stimulation with the same treatments. Using an Integer Linear Programming (ILP) formulation, the proteomic data is combined with a priori knowledge of proteins' connectivity to construct a mechanistic model, predictive of signal transduction in chondrocytes. RESULTS: We were able to validate previous findings regarding major players of cartilage homeostasis and inflammation (e.g., IL1B, TNF, EGF, TGFA, INS, IGF1 and IL6). Moreover, we studied pro-inflammatory mediators (IL1B and TNF) together with pro-growth signals for investigating their role in chondrocytes hypertrophy and highlighted the role of underreported players such as Inhibin beta A (INHBA), Defensin beta 1 (DEFB1), CXCL1 and Flagellin, and uncovered the way they cross-react in the phosphoproteomic level. CONCLUSIONS: The analysis presented herein, leveraged high throughput proteomic data via an ILP formulation to gain new insight into chondrocytes signaling and the pathophysiology of degenerative diseases in articular cartilage.

Comparison of Goldmann and Pascal tonometry in relation to corneal hysteresis and central corneal thickness in nonglaucomatous eyes.

OBJECTIVE: To compare measurements obtained by Goldmann applanation tonometry (GAT) and Pascal dynamic contour tonometry (DCT), and to study their relationship to corneal thickness and biomechanical properties in nonglaucomatous eyes. METHODS: This is a prospective and randomized study of 200 eyes from 200 non-glaucomatous subjects who underwent intraocular pressure (IOP) measurements by GAT and DCT. The two methods were compared and assessed for agreement by means of the Bland-Altman plot. Central corneal thickness (CCT) and corneal hysteresis (CH) were obtained by ultrasound pachymeter and Ocular Response Analyzer, respectively. The effect of CH and CCT was correlated with the DCT/GAT IOP differences. RESULTS: Mean age was 57.4 ± 14.7 years (range 24-82 years). Mean IOP measurements obtained were 16.7 ± 3.2 mmHg by GAT and 19.4 ± 3.3 mmHg by DCT. DCT showed a statistically significant higher mean IOP (2.7 ± 1.9 mmHg, P < 0.001) compared with GAT. Mean CCT and CH were 546.5 ± 40 µm and 10.85 ± 2.0 mmHg, respectively. The differences in IOP (DCT - GAT) were significantly correlated with CCT and CH (Pearson's correlation coefficient r = -0.517 and -0.355, P < 0.0001, respectively). The difference between the two correlation coefficients was statistically significant (P < 0.05, Z-statistic). According to the Bland-Altman plot, the results of the two methods were clinically different. CONCLUSION: Significantly higher IOP readings were obtained by DCT than by GAT in nonglaucomatous subjects. The IOP differences between the two methods were associated with CCT and CH, suggesting that DCT was less dependent on corneal parameters. Each method provides clinically different IOP values, indicating that DCT and GAT should not be used interchangeably.

A method for quantifying cell size from differential interference contrast images: validation and application to osmotically stressed chondrocytes.

An automatic image analysis method was developed to determine the shape and size of spheroidal cells from a time series of differential interference contrast (DIC) images. The program incorporates an edge detection algorithm and dynamic programming for edge linking. To assess the accuracy and working range of the method, results from DIC images of different focal planes and resolutions were compared to confocal images in which the cell membrane was fluorescently labelled. The results indicate that a 1-microm focal drift from the in-focus plane can lead to an overestimation of cell volume up to 14.1%, mostly due to shadowing effects of DIC microscopy. DIC images allow for accurate measurements when the focal plane lies in a zone slightly above the centre of a spherical cell. In this range the method performs with 1.9% overall volume error without taking into account the error introduced by the representation of the cell as a sphere. As a test case, the method was applied to quantify volume changes due to acute changes of osmotic stress.

Hyper-osmotic stress induces volume change and calcium transients in chondrocytes by transmembrane, phospholipid, and G-protein pathways.

Mechanical compression of cartilage is associated with a rise in the interstitial osmotic pressure, which can alter cell volume and activate volume recovery pathways. One of the early events implicated in regulatory volume changes and mechanotransduction is an increase of intracellular calcium ion ([Ca(2+)](i)). In this study, we tested the hypothesis that osmotic stress initiates intracellular Ca(2+) signaling in chondrocytes. Using laser scanning microscopy and digital image processing, [Ca(2+)](i) and cell volume were monitored in chondrocytes exposed to hyper-osmotic solutions. Control experiments showed that exposure to hyper-osmotic solution caused significant decreases in cell volume as well as transient increases in [Ca(2+)](i). The initial peak in [Ca(2+)](i) was generally followed by decaying oscillations. Pretreatment with gadolinium, a non-specific blocker of mechanosensitive ion channels, inhibited this [Ca(2+)](i) increase. Calcium-free media eliminated [Ca(2+)](i) increases in all cases. Pretreatment with U73122, thapsigargin, or heparin (blockers of the inositol phosphate pathway), or pertussis toxin (a blocker of G-proteins) significantly decreased the percentage of cells responding to osmotic stress and nearly abolished all oscillations. Cell volume decreased with hyper-osmotic stress and recovered towards baseline levels throughout the duration of the control experiments. The peak volume change with 550 mOsm osmotic stress, as well as the percent recovery of cell volume, was dependent on [Ca(2+)](i.) These findings indicate that osmotic stress causes significant volume change in chondrocytes and may activate an intracellular second messenger signal by inducing transient increases in [Ca(2+)](i).

Automated standardized pupillometry with optical method for purposes of clinical practice and research.

The aim of the current study was the introduction and standardization of two experimental conditions for dynamic pupillometry. Pupillometry is a method that can provide valuable data concerning the functioning of the autonomous nervous system. The system for recording the pupil reaction was developed in the Laboratory of Clinical Neurophysiology of the 1st Department of Neurology of Aristotle University of Thessaloniki, in co-operation with the Laboratory of Fluid Mechanics of the Aristotle University of Thessaloniki. This system is fully automated. It includes an infra-red video camera, which has the capacity to record in complete darkness, and an SLE (clinical photic stimulator) lamp. A software application automatically performed all the procedures. During the first experiment, one flash was administered. During the second experiment, a series of 25 flashes (1 Hz frequency) was administered. Fifty physically and mentally healthy subjects aged 23-48 years took part in the study. Means, standard deviations and ranges for all variables characterizing normal subjects during both experimental conditions are reported. Test/re-test results and comparisons of the two eyes are also reported. The combined use of these two experimental conditions in dynamic pupillometry may be a very useful tool in medical research. There are already reports on the usefulness of pupillometry in the research of various diseases, including depression and Alzheimer's disease. It is expected that it will also be a valuable research tool in the study of diabetes, alcoholism, myasthenia gravis, cancer, multiple sclerosis, etc.