Deep-Manager: a versatile tool for optimal feature selection in live-cell imaging analysis.

One of the major problems in bioimaging, often highly underestimated, is whether features extracted for a discrimination or regression task will remain valid for a broader set of similar experiments or in the presence of unpredictable perturbations during the image acquisition process. Such an issue is even more important when it is addressed in the context of deep learning features due to the lack of a priori known relationship between the black-box descriptors (deep features) and the phenotypic properties of the biological entities under study. In this regard, the widespread use of descriptors, such as those coming from pre-trained Convolutional Neural Networks (CNNs), is hindered by the fact that they are devoid of apparent physical meaning and strongly subjected to unspecific biases, i.e., features that do not depend on the cell phenotypes, but rather on acquisition artifacts, such as brightness or texture changes, focus shifts, autofluorescence or photobleaching. The proposed Deep-Manager software platform offers the possibility to efficiently select those features having lower sensitivity to unspecific disturbances and, at the same time, a high discriminating power. Deep-Manager can be used in the context of both handcrafted and deep features. The unprecedented performances of the method are proven using five different case studies, ranging from selecting handcrafted green fluorescence protein intensity features in chemotherapy-related breast cancer cell death investigation to addressing problems related to the context of Deep Transfer Learning. Deep-Manager, freely available at https://github.com/BEEuniroma2/Deep-Manager , is suitable for use in many fields of bioimaging and is conceived to be constantly upgraded with novel image acquisition perturbations and modalities.

Machine learning phenomics (MLP) combining deep learning with time-lapse-microscopy for monitoring colorectal adenocarcinoma cells gene expression and drug-response.

High-throughput phenotyping is becoming increasingly available thanks to analytical and bioinformatics approaches that enable the use of very high-dimensional data and to the availability of dynamic models that link phenomena across levels: from genes to cells, from cells to organs, and through the whole organism. The combination of phenomics, deep learning, and machine learning represents a strong potential for the phenotypical investigation, leading the way to a more embracing approach, called machine learning phenomics (MLP). In particular, in this work we present a novel MLP platform for phenomics investigation of cancer-cells response to therapy, exploiting and combining the potential of time-lapse microscopy for cell behavior data acquisition and robust deep learning software architectures for the latent phenotypes extraction. A two-step proof of concepts is designed. First, we demonstrate a strict correlation among gene expression and cell phenotype with the aim to identify new biomarkers and targets for tailored therapy in human colorectal cancer onset and progression. Experiments were conducted on human colorectal adenocarcinoma cells (DLD-1) and their profile was compared with an isogenic line in which the expression of LOX-1 transcript was knocked down. In addition, we also evaluate the phenotypic impact of the administration of different doses of an antineoplastic drug over DLD-1 cells. Under the omics paradigm, proteomics results are used to confirm the findings of the experiments.

LOX-1 and cancer: an indissoluble liaison.

Recently, a strong correlation between metabolic disorders, tumor onset, and progression has been demonstrated, directing new therapeutic strategies on metabolic targets. OLR1 gene encodes the LOX-1 receptor protein, responsible for the recognition, binding, and internalization of ox-LDL. In the past, several studied, aimed to clarify the role of LOX-1 receptor in atherosclerosis, shed light on its role in the stimulation of the expression of adhesion molecules, pro-inflammatory signaling pathways, and pro-angiogenic proteins, including NF-kB and VEGF, in vascular endothelial cells and macrophages. In recent years, LOX-1 upregulation in different tumors evidenced its involvement in cancer onset, progression and metastasis. In this review, we outline the role of LOX-1 in tumor spreading and metastasis, evidencing its function in VEGF induction, HIF-1alpha activation, and MMP-9/MMP-2 expression, pushing up the neoangiogenic and the epithelial-mesenchymal transition process in glioblastoma, osteosarcoma prostate, colon, breast, lung, and pancreatic tumors. Moreover, our studies contributed to evidence its role in interacting with WNT/APC/ß-catenin axis, highlighting new pathways in sporadic colon cancer onset. The application of volatilome analysis in high expressing LOX-1 tumor-bearing mice correlates with the tumor evolution, suggesting a closed link between LOX-1 upregulation and metabolic changes in individual volatile compounds and thus providing a viable method for a simple, non-invasive alternative monitoring of tumor progression. These findings underline the role of LOX-1 as regulator of tumor progression, migration, invasion, metastasis formation, and tumor-related neo-angiogenesis, proposing this receptor as a promising therapeutic target and thus enhancing current antineoplastic strategies.

Accelerating the experimental responses on cell behaviors: a long-term prediction of cell trajectories using Social Generative Adversarial Network.

The incremented uptake provided by time-lapse microscopy in Organ-on-a-Chip (OoC) devices allowed increased attention to the dynamics of the co-cultured systems. However, the amount of information stored in long-time experiments may constitute a serious bottleneck of the experimental pipeline. Forward long-term prediction of cell trajectories may reduce the spatial-temporal burden of video sequences storage. Cell trajectory prediction becomes crucial especially to increase the trustworthiness in software tools designed to conduct a massive analysis of cell behavior under chemical stimuli. To address this task, we transpose here the exploitation of the presence of "social forces" from the human to the cellular level for motion prediction at microscale by adapting the potential of Social Generative Adversarial Network predictors to cell motility. To demonstrate the effectiveness of the approach, we consider here two case studies: one related to PC-3 prostate cancer cells cultured in 2D Petri dishes under control and treated conditions and one related to an OoC experiment of tumor-immune interaction in fibrosarcoma cells. The goodness of the proposed strategy has been verified by successfully comparing the distributions of common descriptors (kinematic descriptors and mean interaction time for the two scenarios respectively) from the trajectories obtained by video analysis and the predicted counterparts.

Discovering the hidden messages within cell trajectories using a deep learning approach for in vitro evaluation of cancer drug treatments.

We describe a novel method to achieve a universal, massive, and fully automated analysis of cell motility behaviours, starting from time-lapse microscopy images. The approach was inspired by the recent successes in application of machine learning for style recognition in paintings and artistic style transfer. The originality of the method relies i) on the generation of atlas from the collection of single-cell trajectories in order to visually encode the multiple descriptors of cell motility, and ii) on the application of pre-trained Deep Learning Convolutional Neural Network architecture in order to extract relevant features to be used for classification tasks from this visual atlas. Validation tests were conducted on two different cell motility scenarios: 1) a 3D biomimetic gels of immune cells, co-cultured with breast cancer cells in organ-on-chip devices, upon treatment with an immunotherapy drug; 2) Petri dishes of clustered prostate cancer cells, upon treatment with a chemotherapy drug. For each scenario, single-cell trajectories are very accurately classified according to the presence or not of the drugs. This original approach demonstrates the existence of universal features in cell motility (a so called "motility style") which are identified by the DL approach in the rationale of discovering the unknown message in cell trajectories.

The influence of spatial and temporal resolutions on the analysis of cell-cell interaction: a systematic study for time-lapse microscopy applications.

Cell-cell interactions are an observable manifestation of underlying complex biological processes occurring in response to diversified biochemical stimuli. Recent experiments with microfluidic devices and live cell imaging show that it is possible to characterize cell kinematics via computerized algorithms and unravel the effects of targeted therapies. We study the influence of spatial and temporal resolutions of time-lapse videos on motility and interaction descriptors with computational models that mimic the interaction dynamics among cells. We show that the experimental set-up of time-lapse microscopy has a direct impact on the cell tracking algorithm and on the derived numerical descriptors. We also show that, when comparing kinematic descriptors in two diverse experimental conditions, too low resolutions may alter the descriptors' discriminative power, and so the statistical significance of the difference between the two compared distributions. The conclusions derived from the computational models were experimentally confirmed by a series of video-microscopy acquisitions of co-cultures of unlabelled human cancer and immune cells embedded in 3D collagen gels within microfluidic devices. We argue that the experimental protocol of acquisition should be adapted to the specific kind of analysis involved and to the chosen descriptors in order to derive reliable conclusions and avoid biasing the interpretation of results.

Structural investigation of nucleophosmin interaction with the tumor suppressor Fbw7γ.

Nucleophosmin (NPM1) is a multifunctional nucleolar protein implicated in ribogenesis, centrosome duplication, cell cycle control, regulation of DNA repair and apoptotic response to stress stimuli. The majority of these functions are played through the interactions with a variety of protein partners. NPM1 is frequently overexpressed in solid tumors of different histological origin. Furthermore NPM1 is the most frequently mutated protein in acute myeloid leukemia (AML) patients. Mutations map to the C-terminal domain and lead to the aberrant and stable localization of the protein in the cytoplasm of leukemic blasts. Among NPM1 protein partners, a pivotal role is played by the tumor suppressor Fbw7γ, an E3-ubiquitin ligase that degrades oncoproteins like c-MYC, cyclin E, Notch and c-jun. In AML with NPM1 mutations, Fbw7γ is degraded following its abnormal cytosolic delocalization by mutated NPM1. This mechanism also applies to other tumor suppressors and it has been suggested that it may play a key role in leukemogenesis. Here we analyse the interaction between NPM1 and Fbw7γ, by identifying the protein surfaces implicated in recognition and key aminoacids involved. Based on the results of computational methods, we propose a structural model for the interaction, which is substantiated by experimental findings on several site-directed mutants. We also extend the analysis to two other NPM1 partners (HIV Tat and CENP-W) and conclude that NPM1 uses the same molecular surface as a platform for recognizing different protein partners. We suggest that this region of NPM1 may be targeted for cancer treatment.

Systematic approach in Mg2+ ions analysis with a combination of tailored fluorophore design.

A systematic study of a series of diaza-18-crown-6 8-hydroxyquinoline (DCHQ) chemosensors, devoted to Mg(II) ion detection, was performed. Functionalization of DCHQ by peripheral substituents allowed the development of novel all-solid-state optodes via inclusion inside PVC-based polymeric films. The influence on the DCHQ-based optode response of the lipophilic sites functionalization and of the nature of the plasticizer, was investigated. Fluorimetric studies on optodes sensitivity towards a number of different metal cations (Ca2+, Na+, K+, Li+, Co2+, Cd2+, Pb2+, Cu2+, Hg2+, Zn2+) and NH4+ were carried out. The results demonstrated the suitability of the DCHQ optodes to perform fast monitoring (<10s) of magnesium (II) ions. Emission light signal was sufficiently brilliant to be captured by a low-cost computer webcam. The phenyl-substituted DCHQ-Ph derivative showed the best performance with a wide range for Mg(II) ion determination between 2.7 × 10-7 and 2.2 × 10-2 mol/L. It was possible, therefore, to determine the concentrations of Mg(II) in commercial fertilizer samples by DCHQ-Ph-based optodes with acceptable results: recoveries of 96.2-104.9% and relative standard deviation (RSD, n = 6) less than 5%. Moreover, in comparison to single sensors, the use of an array composed of five optodes (the ones showing the best performances in the preliminary tests) has allowed to reduce the RSD of magnesium determination in real samples (down to 3.7% with respect to 5.5% for single optodes) and to achieve a detection limit (estimated by s/n = 3 method) as low as 4.6 × 10-7 mol/L.

Vortexes tune the chirality of graphene oxide and its non-covalent hosts.

Graphene oxide (GO) is one of the most appealing bidimensional materials able to interact non-covalently with achiral molecules and to act as chiral inducers. Vortexes can tune chirality and, consequently transfer a specific handedness to non-covalent host molecules, either when dispersed in water or when deposited on a solid surface.

Electronic tongue for microcystin screening in waters.

The potentiometric E-tongue system was employed for water toxicity estimation in terms of cyanobacterial microcystin toxins (MCs) detection. The data obtained from E-tongue were correlated to the MCs content detected by the standard chromatographic technique UHPLC-DAD (Ultra High Performance Liquid Chromatography with Diode Array Detector), as far as by the colorimetric enzymatic approach. The prediction of MCs released by toxic Microcystis aeruginosa strains was possible with Root Mean Squared Error of Validation (RMSEV) lower or very close to 1µg/L, the provisional guideline value of WHO for MCs content in potable waters. The application of E-tongue system opens up a new perspective offset for fast and inexpensive analysis in the field of environmental monitoring, offering also the possibility to distinguish toxin producing and non-toxic M. aeruginosa strains present in potable water.

Disseminated, large-sized neonatal pneumatoceles: the wait-and-see strategy.

Pulmonary pneumatoceles are thin-walled, air-filled cysts that develop within the lung parenchyma. Most often, they occur as a sequel of acute pneumonia, commonly caused by Staphylococcus aureus in children. Limited data are available about infective pulmonary cysts in newborns. We report a case of a newborn, who developed multiple pneumatoceles after Escherichia coli pneumonia.

From peptides to small molecules: an intriguing but intricated way to new drugs.

A variety of peptides active in biological pathways have been identified e.g. receptor antagonists or inhibitors of protein-protein interactions and several peptide or peptide-derived compounds are on the drug market or in clinical trials. Through the rational design or the combinatorial preparation and High-throughput screening of arrays of compounds, peptides play a pivotal role for the rapid identification of ligands, but, despite these favorable properties, they often present poorer bioavailability and lower metabolic stability respect to traditional drugs. The process of conversion of a peptide in a small molecule provides the reduction of the peptide to the minimum active sequence (MAS) testing truncated peptides from the C- and N- termini alternatively. Then the influence of individual amino acid on the biological activity is determined by systematically replacing each residue in the peptide with specific amino acids. After structure-activity relationship (SAR) of each amino acid in the sequence has been assessed, the bioactive conformational flexibility is reduced by introducing constraints at various positions. These features are used for the design of a pharmacophore model in which functional groups crucial for activity are pre-positioned. Here we propose a panoramic review of the common principles for the conversion of peptides into small organic molecules and the most interesting findings in peptide-based leads of the last decades.

Gold nanoparticles-peptide based gas sensor arrays for the detection of food aromas.

A gas sensor array based on peptide modified gold nanoparticles deposited onto 20MHz quartz crystal microbalances has been realized. Glutathione and its constituting aminoacids and dipeptides have been used as ligands. A great increase in sensitivity (2 orders of magnitude) was achieved using gold nanoparticles versus monolayer modified QCMs. The sensors have been characterised in terms of sensitivity for hexane, water, trimethylammine and ethanol. Highest sensitivity was found for water. The ability to discriminate typical food aromas as cis-3-hexenol, isopentylacetate, ethylacetate, and terpinen-4-ol dissolved in different solvents was studied using a gas sensor array constituted by gold nanoparticles modified with the glutathione peptides, thioglycolic acid and an heptapeptide. The array was found able to discriminate the food aromas, the response being dependent on the polarity of the solvent used. Tests on real olive oil samples gave a satisfactory separation among samples having defects versus non defected samples demonstrating that this approach has high potential for the development of gas sensor arrays to be used in real samples.

In situ detection of lung cancer volatile fingerprints using bronchoscopic air-sampling.

Lung cancer diagnosis via breath analysis has to overcome some issues that can be summarized by two crucial points: (1) further developments for more performant breath sampling technologies; (2) discovering more differentiated volatile fingerprints to be ascribed to specific altered biological mechanisms. The present work merges these two aspects in a pilot study, where a breath volume, sampled via endoscopic probe, is analyzed by an array of non-selective gas sensors. Even if the original non-invasive methods of breath analysis has been laid in favour of the endoscopic means, the innovative technique here proposed allows the analysis of the volatile mixtures directly sampled near the tumor mass. This strategy could open the way for a better understanding of the already obtained discrimination among positive and negative cancer cases. The results obtained so far confirm the established discrimination capacity. This allows to discriminate the different subtypes of lung cancer with 75% of correct classification between adenocarcinoma and squamous cell carcinoma. This result suggests that a 'zoom-in' on the cancer settled inside the human body can increase the resolution power of key-volatiles detection, allowing the discrimination among different cancer fingerprints. We report this novel technique as a robust support for a better comprehension of the promising results obtained so far and present in literature; it is not to be intended as a replacement for non-invasive breath sampling procedure with the endoscope.

Carbon nanotube films as a platform to transduce molecular recognition events in metalloporphyrins.

Porphyrins have been widely used for many years as functional materials for chemical sensors. Their outstanding chemical features are balanced by some restrictions in terms of transduction techniques. In particular, porphyrin layers are barely conductive, with the consequence that the fabrication of porphyrin based chemiresistors is not possible, except in few rare cases. On the other hand, carbon nanotubes (CNTs) have superior electric properties ranging from metallic to semiconductor in character. Although the conductivity of CNTs is very sensitive to adsorbed molecules, it should be considered that the adsorption onto carbon structures is also scarcely selective and cannot be modified unless other molecular recognition systems are coupled with the CNTs. Following this approach, in this paper we investigated the sensing properties of hybrid CNT-porphyrin films to explore the possibility of transducing the adsorption events occurring in a porphyrin layer into resistance changes of the CNT layers. The results obtained indicate that the presence of the porphyrin films increases the sensitivity of the electric resistance of the CNTs to the concentration of volatile compounds. This enhancement is probably due to the catalytic effect of the metalloporphyrin in conveying the charge transfer from the adsorbate molecule to the CNTs substrate. This property of metalloporphyrins may introduce a further differentiation between porphyrin based sensors that could be positively utilized in sensor array configurations.

Identification of melanoma with a gas sensor array.

BACKGROUND/PURPOSE: The relationship between diseases and alterations of the airborne chemicals emitted from the body has been found in many different pathologies and in particular for various forms of cancer. Metabolism of cancer cells is greatly altered during their lifetime; then, modification of chemicals is supposed to be large around cancer tissues. Positive hints in this direction were provided, as an example, on studying the breath composition of lung cancer-affected subjects. Besides the conventional analytical approaches, in recent years sensor arrays were also applied to these researches considering the chemical composition changes as those occurring in other applications such as for instance, those dealing with food quality measurements. METHODS: In this paper, the first application of sensor arrays to study the differentiation between melanomas and nevi, namely malignant and benign affection of melanocytary cells, respectively, is presented and discussed. The localization of lesions on the skin surface made possible the utilization of differential measurements aimed at capturing the differences between two adjacent skin regions. This approach strongly reduces the influence of skin headspace variability due to the peculiar subjective odour background and the skin odour variability. The measurement campaign involved 40 cases; 10 of these were diagnosed melanomas referred to surgical intervention. Nine of these diagnoses were further confirmed by histological examinations of the removed tissue and one was a false positive. RESULTS: The differences in the chemical composition of headspace were verified with a gas-chromatographic investigation, and the classification of electronic nose data provided an estimated cross-validated accuracy of the same order of magnitude as the currently used diagnostic instruments. CONCLUSION: Electronic nose sensors have been shown to have good sensitivity towards volatile organic compounds emitted by skin lesions, and the method seems to be effective for malign lesions identification. The results presented in this paper encourage a second experimental campaign with a larger number of participants and a systematic use of gas chromatography mass spectrometer technology in order to identify some possible melanoma biomarkers.

Direct and two-stage data analysis procedures based on PCA, PLS-DA and ANN for ISE-based electronic tongue-Effect of supervised feature extraction.

A novel strategy of data analysis for artificial taste and odour systems is presented in this work. It is demonstrated that using a supervised method also in feature extraction phase enhances fruit juice classification capability of sensor array developed at Warsaw University of Technology. Comparison of direct processing (raw data processed by Artificial Neural Network (ANN), raw data processed by Partial Least Squares-Discriminant Analysis (PLS-DA)) and two-stage processing (Principal Components Analysis (PCA) outputs processed by ANN, PLS-DA outputs processed by ANN) is presented. It is shown that considerable increase of classification capability occurred in the case of the new method proposed by the authors.

Piezoelectric sensors for dioxins: a biomimetic approach.

The aim of this work was to design a fast, cheap and easy to use analytical system for dioxins. Piezoelectric sensors coupled with the pentapeptides as biomimetic traps (the receptors), selective for the dioxins, were used for the realisation of this analytical system. A methodology to select specific receptors among all possible pentapeptides randomly generated was represented by the use of molecular modelling software. Three peptides called later on A, B and C (A:[N]Asn-Phe-Gln-Gly-Ile[C]; B:[N]Asn-Phe-Gln-Gly-Gln[C]; C:[N]Asn-Phe-Gln-Gly-Phe[C]), were selected and evaluated for their potential usage as artificial receptors in solid-gas analysis by using a quartz crystal microbalance (QCM) sensors array. The peptide sequences were functionalised by two terminal cysteine residues in order to achieve a covalent interaction with the QCM gold surface. A manganese-porphyrin complex and two other pentapeptides, a pentaglutamine (pentapeptide D) and a pentalysine (pentapeptide E), were used as negative control sensors. The QCM sensors (A, B and C) gave a good linearity against different sample concentrations of the 2,3,7,8-tetrachlorinated dibenzo-p-dioxin (TCDD) and a mixture of dioxins. In particular, the selectivity against 2,3,7,8-TCDD was nicely correlated to the estimated binding energy of the receptors calculated by computational modelling. The cross-reactivity of the system was quantified using commercial polychlorinated biphenyls (PCBs) mixtures (dioxin-like compounds).

Monitoring of biological odour filtration in closed environments with olfactometry and an electronic nose.

Air treatment with a compact biological membrane filter, and air quality monitoring with an electronic nose were tested in the laboratory on air from a cage containing six mice. Additional analyses of air to and from the filter were performed using olfactometry and ammonia and hydrogen sulphide gas detection tubes. The biological air filter is a module containing biofilm-coated membrane fibres that separate a closed liquid loop from a gas phase. Odour compounds and oxygen diffuse through the membranes from the gas phase to the biofilm, where they are degraded to carbon dioxide and water. The prototype "ENQBE" electronic nose is based on an array of eight thickness shear mode resonators (TSMR), also known in the literature as quartz microbalance sensors. The chemical sensitivity is given by molecular films of metalloporphyrins and similar compounds. Chemical interaction of compounds in the air with the vibrating sensors induces a frequency change of the vibration that can be measured as a signal. The air from the mouse cage had a strong odour (3490 OUE/m3). The biological membrane filter performed well, achieving over 80% odour and ammonia reduction. The electronic nose signal could be correlated with the inlet and outlet air-quality of the biological filter, making it a promising method for monitoring air quality in closed environments.