Call for Late-Breaking Abstracts


Experimental determination and data-driven prediction of homotypic transmembrane domain interfaces

Dieter Langosch

Lehrstuhl für Chemie der Biopolymere, Technische Universität München, Freising, Germany 

Yao Xiao

Lehrstuhl für Chemie der Biopolymere,Technische Universität München, Freising, Germany

Bo Zeng 

Department of Bioinformatics, Technische Universität München, Freising, Germany

Nicola Brenner

Lehrstuhl für Chemie der Biopolymere, Technische Universität München, Freising, Germany

Dmitrij Frishman 

Department of Bioinformatics, Technische Universität München, Freising, Germany

Marc Teese

Lehrstuhl für Chemie der Biopolymere, Technische Universität München, Freising, Germany

Interactions between their transmembrane domains (TMDs) frequently support the assembly of single-pass membrane proteins to non-covalent complexes. Yet, the TMD- TMD interactome remains largely uncharted. With a view to predicting homotypic TMD-TMD interfaces from primary structure, we performed a systematic analysis of their physical and evolutionary properties. To this end, we generated a dataset of 54 self-interacting TMDs. This dataset contains interfaces of nine TMDs from bitopic human proteins (Ire1, Armcx6, Tie1, ATP1B1, PTPRO, PTPRU, PTPRG, DDR1, and Siglec7) that were experimentally identified here and combined with literature data. We show that interfacial residues of these homotypic TMD-TMD interfaces tend to be more conserved, coevolved and polar than non-interfacial residues. Further, we suggest for the first time that interface positions are deficient in β-branched residues, and likely to be located deep in the hydrophobic core of the membrane. Overrepresentation of the GxxxG motif in interfaces is strong, but that of (small)xxx(small) motifs is weak. The multiplicity of these features and the individual character of TMD-TMD interfaces, as uncovered here, prompted us to train a machine learning algorithm. The resulting prediction method, THOIPA, excels in the prediction of interface residues from evolutionary sequence data.


A Data-Driven Approach to Assess the Role of Neuroblastoma-Derived Exosomes in Cancer Dissemination

Luca Zanella

Dept. of Industrial Engineering (DII), University of Padova

Pina Fusco

– Dept. of Industrial Engineering (DII), University of Padova

– NBTECH Laboratory, Fondazione Istituto di Ricerca Pediatrica Citta` della Speranza

Pierantonio Facco

Dept. of Industrial Engineering (DII), University of Padova

Fabrizio Bezzo

Dept. of Industrial Engineering (DII), University of Padova

Elisa Cimetta

– Dept. of Industrial Engineering (DII), University of Padova

– NBTECH Laboratory, Fondazione Istituto di Ricerca Pediatrica Citta` della Speranza

Neuroblastoma is a heterogeneous pediatric malignancy, originating from progenitor cells of the sympathetic nervous system, that accounts for 8-10% of all childhood cancers. Metastases are found in approximately half of the patients, typically located in bone, bone marrow, lymph nodes, liver and skin. Neuroblastoma tendency to metastasize is due to the communication with nearby cells and distant organs. The regulation of tumor progression is strongly affected by microenvironmental cues. 

Exosomes are extracellular vesicle-like structures, ranging in size between 40 and 100 nm. Exosomal cargo includes micro RNAs (miRNAs), proteins, growth factors and nucleic acids [1]. Tumor secreted exosomes deliver signals that act by regulating cell-cell communication, promoting tumor progres- sion, invasion and metastasis. 

Hypoxia is a key feature of solid tumors, known to (a) favor NB metastasis and dedifferentiation towards immature stem cell-like phenotypes and to (b) stimulate release of exosomes, that ultimately facilitate intercellular communication at distant sites. 

The objective of this work is to characterize the miRNAs and proteomic cargo of exosomes secreted by two Neu- roblastoma cell lines, SN-K-AS and SN-K-DZ, cultured in atmospheres at different oxygenation conditions, to identify an exosomal signature associated to metastatic dissemination. This is a preliminary analysis that paves the ground to the investigation of the combined effect of exosomes concentration gradients and oxygen levels on target cells. 

Two datasets have been analysed. The first comprises the expression values of 250 miRNAs (number of samples 6), the latter those of 279 proteins (number of samples 18, three biological replicates for each cell line at the given conditions). miRNA and Proteome cargo profiles have been analyzed by FirePlex Discovery Panel and mass spectrometry, respectively. Each observation represents exosomes isolated from either SN- K-AS or SN-K-DZ cultured in one of the three oxygenation conditions, namely normoxia (20% O2, on a mole basis), hypoxia (1.5% O2) and reoxygenation (cells cultured for 24 h at 1.5% O2, followed by 24 h at 20% O2). A list of miRNAs and proteins that are differentially expressed in 1. each cell line and 2. each oxygenation level have been identified by Principal Components Analysis (PCA) [2] and Projection to 

Latent Structures by Partial Least Squares for Discriminant Analysis (PLS-DA) [3]. 

PCA applies singular values decomposition to a covariance matrix, finding the directions of maximum variability within the dataset. The method produces scores and loadings plots, whose inspection is used to assess samples and variables correlations. 

PLS is a linear regression technique, where regressors and response variables are projected into a new space. The use of categorical arrays as response variables and a class assignment rule allows using this method for classification (PLS-DA). 

Standardization has been employed as the preprocessing method in all the analyses. A three-components PCA model calibrated on the miRNA dataset captures 90% of the original variables. The variability within the dataset is mainly ascribed to differences between the two cell lines, rather than the oxygenation conditions. However, there exists one principal component (variance captured 6.95%) that allows the separa- tion based on the different oxygenation conditions. A list of 45 miRNAs that characterize the three oxygenation conditions in both cell lines has been identified, with 15 miRNAs highly expressed in each of them. These miRNAs have been chosen as the 6% with the highest loading on this component (hypoxia), around zero (reoxygenation) and lowest loading (normoxia). 

Analogously, a list of proteins highly expressed in the three oxygenation conditions has been found by calibrating separate PCA models on the samples from the two cell lines. In both cases, variables have been compared to those selected by inspection of the weights plot in a PLS-DA model. 

These preliminary results will be subject to validation. They will represent the starting point for the identification of predictive biomarkers for metastatic spread in Neuroblastoma. 


[1] Braicu et al., Cell Death Differ., 2015. 

[2] Wold et al., Chemom. Intell. Lab. Syst., 1987.

[3] Geladi et al., Anal. Chim. Acta, 1986.


A Bi-clustering of metabolic data and an R package for metabolic bicluster analysis

Quan Gu 

University of Glasgow

Jonathan Lim

University of Glasgow

Metabolic phenotyping technologies based on Nuclear Magnetic Spectroscopy (NMR) and Mass Spectrometry (MS) generate vast amounts of unrefined data from biological samples. The proposed strategy incorporates bi-cross validation and statistical segmentation techniques to automatically determine the number and structure of bi-clusters. We perform the comparative unsupervised learning analysis of the proposed strategy with other bi-clustering approaches, which were developed in the context of genomics and transcriptomics research. We demonstrate the superior performance of the proposed bi-clustering strategy on the real bacterial metabolic data. Here we also develop mfBiclust, an R package with a Shiny-based GUI that enables users to apply recently developed biclustering pipelines to biological datasets. Visualization and export of the results facilitates functional characterization of the observed biclusters and makes mfBiclust potentially useful for analysing any -omics assay. 


A 5G monitoring system through wearable sensors and machine learning for personalized medicine

Antonio Pallotti, Emanuela Tagliente, Leandro Lucangeli

Technocience – University of Rome San Raffaele 

In the clinical field, from 2010 to 2016, there was an increase in the global telemedicine research, mainly due to the increase in remote monitoring of patient healthcare [1]; it is suggested to increase a more frequent evaluation of the patient health state [2], with the opportunity to remotely modify the medical management. For chronic diseases, which are more and more rising, the present work wants to provide a reduction in acute hospitalization [3]. The monitoring of physiological symptoms and physiological signals can allow the development of predictive models and the development of the most appropriate treatment [4] [5]; New statistical and computational methods have been implemented, based on machine learning, for the development of predictive models able to distinguish in risk factor classes [4]. Getting to store a greater amount of data [6] on the patient’s health allows these predictive models to be better trained and validated. The model result is submitted to the doctor’s consultation, in order to proceed in the administration of personalized therapy as best as possible. It’s necessary to monitor the clinical course, especially in certain diseases [7]. This creates the basis for the development of teleassistance (in the field of telemedicine [8] [9]), seen as the set of techniques that represent the Information and communication technology (ICT), which are applied in medical practice [10]. This set of applications should contribute significantly to improve the efficiency of the patient healtcare service. The present work has as objectives[10]: a) the time interval decreasing between diagnosis and therapy; (b) the safely registration of the clinical history for each patient(using the electronic health record – EHR), so that can be easily evaluated by other experts; (c) the decreasing of redundant diagnostic tests; (d) the prevention of unnecessary patient transportation, who remains in the home environment keeping on therapy. The need for developing accurate models, requires a considerable amount of reliable data on the patient’s state of health, and that can be satisfied by the use of wearable sensors that work remotely, through 5G network, and automatically with respect to the attending physician, who can consult in any moment the data sent to the cloud platform, which acting as a database [11]. The continuous advancement of sensor technologies, integrated systems, wireless communication systems, nanotechnologies has made it possible to develop intelligent systems for the continuous human activity recognition (HAR) [12]. The wearable sensors used in this work are the ones for the following physiological investigations: Electrokinesiology (EKG), Electromyography (EMG), Electrocardiography (ECG), Electrooculography (EOG), pedobarography. In addition, there are also sensors for metabolic activity detection. 


[1] Silva, et al., J. Biomed. Inform., 2015. 

[2] Jayaram et al. CircOutcomes, 2017. 

[3] Paré et al., J. Biomed. Inform., 2007.

[4] Orchard et al., J.Med. Internet Res., vol.20, no.9 

[5] Zhao et al., PLoS One, 2017. 

[6] Manogaran, et al, Future Gener. Comput. Syst., 2018. 

[7] Chaudhry et al., N. Engl. J. Med., 2010. 

[8] M. Young, The Technical Writer’s Handbook. Mill Valley, CA: University Science, 1989. 

[9] Ministero della Salute. TELEMEDICINA Linee di indirizzo nazionali. In: Conferenza Stato Regioni. Atti n. 2012 

[10] Pelletier-Fleuryet al., Health Policy, 1999. 

[11] Blasco, et al., J. Cardiopul. Rehabil., 2012. 

[12] Mukhopadhyay, S.C., IEEE sensors journal, 2014. 


Image Object Detection Facilitates the Study of Fucosylation in Multicellular Tumour Spheroids

Regina-Veronicka Kalaydina

Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, Canada

Alexandra Wojaczek 

School of Computing, Queen’s University, Kingston, Canada

Mohammed Gasmallah 

School of Computing, Queen’s University, Kingstone, Canada

Hedi Zhou 

Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, Canada 

Dr. Farhana Zulkernine 

School of Computing, Queen’s University, Kingston, Canada 

Dr. Myron R. Szewczuk 

Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, Canada 

The process of tumor formation remains poorly understood. Glycosylation likely plays a critical role; however, little is known about fucosylation, a biochemical event involving the addition of fucose sugars. Three-dimensional (3D) models of cancer are commonly utilized to gain insight into tumor formation as they mimic the chemical and physical properties of tumors. Multicellular tumor spheroids (MTS) are the prevailing 3D model, whereby favorable cell culture conditions induce cell aggregation into spherical masses (spheroids). Spheroids undergo phase-contrast microscopy imaging and manual identification and measurement. However, this approach is time- consuming, subject to bias, and a lack of standardized analysis automation protocols hinders translation. Here, the role of fucosylation, which has been linked to aggressive prostate cancers, was investigated in spheroid formation using DU145 prostate cancer cells. Alpha 1,6 and 1,2 fucose linkages present in DU145 cells were blocked with lectins prior to addition of a peptide that causes spheroid formation. Images of spheroids were analyzed manually and automatically via the implementation of Darkflow YOLOv2, a single-phase object detector possessing a 24- layer convolutional neural network. Previous work in our lab has shown that spheroids are bound and detected with an F1 score of 76%, IOU of 69%, and 3.99% average error in volume estimation. We found that automated image object detection was not inferior to a manual approach in detection and measurement of DU145 prostate cancer spheroids treated with lectins. We have also shown that fucosylation is an essential event in spheroid formation, as confirmed by our image object detection approach. Understanding the contribution of fucosylation to spheroid formation may give insight into in vivo tumor formation. Moreover, we highlight an approach to improve the high throughput screening potential of MTS. 


Call for Late-Breaking Abstracts

One-page abstracts describing late-breaking developments within the scope of the conference are solicited for presentation at the Poster Snapshot Session and the Poster Session. Abstracts will be published on the conference website.

Authors of accepted abstracts will individually retain copyright (and all other rights) to their abstracts. Accepted abstracts with no author registered by the deadline will not appear on the conference website.

Abstracts should be of the same format as full papers (see formatting details) and submitted using the same link as full papers before the 31th of May.

All abstracts must be submitted via the EasyChair online submission system.