ERC FUNDED PROJECTS

Computer animation has traditionally been associated with applications in virtual-reality-based training, video games or feature films. However, interactive animation is gaining relevance in a more general scope, as a tool for early-stage analysis, design and planning in many applications in science and engineering. The user can get quick and visual feedback of the results, and then proceed by refining the experiments or designs. Potential applications include nanodesign, e-commerce or tactile telecommunication, but they also reach as far as, e.g., the analysis of ecological, climate, biological or physiological processes. The application of computer animation is extremely limited in comparison to its potential outreach due to a trade-off between accuracy and computational efficiency. Such trade-off is induced by inherent complexity sources such as nonlinear or anisotropic behaviors, heterogeneous properties, or high dynamic ranges of effects. The Animetrics project proposes a modeling and animation methodology, which consists of a multi-scale decomposition of complex processes, the description of the process at each scale through combination of simple local models, and fitting the parameters of those local models using large amounts of data from example effects. The modeling and animation methodology will be explored on specific problems arising in complex mechanical phenomena, including viscoelasticity of solids and thin shells, multi-body contact, granular and liquid flow, and fracture of solids.

1 277 969 €

Start date: 2012-01-01, End date: 2016-12-31

The goal of the project is to automatically analyse human activities observed in videos. Any solution to this problem will allow the development of novel applications. It could be used to create short videos that summarize daily activities to support patients suffering from Alzheimer's disease. It could also be used for education, e.g., by providing a video analysis for a trainee in the hospital that shows if the tasks have been correctly executed. The analysis of complex activities in videos, however, is very challenging since activities vary in temporal duration between minutes and hours, involve interactions with several objects that change their appearance and shape, e.g., food during cooking, and are composed of many sub-activities, which can happen at the same time or in various orders. While the majority of recent works in action recognition focuses on developing better feature encoding techniques for classifying sub-activities in short video clips of a few seconds, this project moves forward and aims to develop a higher level representation of complex activities to overcome the limitations of current approaches. This includes the handling of large time variations and the ability to recognize and locate complex activities in videos. To this end, we aim to develop a unified model that provides detailed information about the activities and sub-activities in terms of time and spatial location, as well as involved pose motion, objects and their transformations. Another aspect of the project is to learn a representation from videos that is not tied to a specific source of videos or limited to a specific application. Instead we aim to learn a representation that is invariant to a perspective change, e.g., from a third-person perspective to an egocentric perspective, and can be applied to various modalities like videos or depth data without the need of collecting massive training data for all modalities. In other words, we aim to learn the essence of activities.

1 499 875 €

Start date: 2016-06-01, End date: 2021-05-31

ATM is the protein kinase that is mutated in the hereditary autosomal recessive disease ataxia telangiectasia (A-T). A-T patients display immune deficiencies, cancer predisposition and radiosensitivity. The molecular role of ATM is to respond to DNA damage by phosphorylating its substrates, thereby promoting repair of damage or arresting the cell cycle. Following the induction of double-strand breaks (DSBs), the NBS1 protein is required for activation of ATM. But ATM can also be activated in the absence of DNA damage. Treatment of cultured cells with hypotonic stress leads to the activation of ATM, presumably due to changes in chromatin structure. We have recently described a second ATM cofactor, ATMIN (ATM INteractor). ATMIN is dispensable for DSBs-induced ATM signalling, but ATM activation following hypotonic stress is mediated by ATMIN. While the biological role of ATM activation by DSBs and NBS1 is well established, the significance, if any, of ATM activation by ATMIN and changes in chromatin was up to now completely enigmatic. ATM is required for class switch recombination (CSR) and the suppression of translocations in B cells. In order to determine whether ATMIN is required for any of the physiological functions of ATM, we generated a conditional knock-out mouse model for ATMIN. ATM signaling was dramatically reduced following osmotic stress in ATMIN-mutant B cells. ATMIN deficiency led to impaired CSR, and consequently ATMIN-mutant mice developed B cell lymphomas. Thus ablation of ATMIN resulted in a severe defect in ATM function. Our data strongly argue for the existence of a second NBS1-independent mode of ATM activation that is physiologically relevant. While a large amount of scientific effort has gone into characterising ATM signaling triggered by DSBs, essentially nothing is known about NBS1-independent ATM signaling. The experiments outlined in this proposal have the aim to identify and understand the molecular pathway of ATMIN-dependent ATM signaling.

1 499 881 €

Start date: 2012-02-01, End date: 2018-01-31

Genetic sequences have had an enormous impact on our understanding of biology. The expectation is that biological network data will have a similar impact. However, progress is hindered by a lack of sophisticated graph theoretic tools that will mine these large networked datasets. In recent breakthrough work at the boundary of computer science and biology supported by my USA NSF CAREER award, I developed sensitive network analysis, comparison and embedding tools which demonstrated that protein-protein interaction networks of eukaryotes are best modeled by geometric graphs. Also, they established phenotypically validated, unprecedented link between network topology and biological function and disease. Now I propose to substantially extend these preliminary results and design sensitive and robust network alignment methods that will lead to uncovering unknown biology and evolutionary relationships. The potential ground-breaking impact of such network alignment tools could be parallel to the impact the BLAST family of sequence alignment tools that have revolutionized our understanding of biological systems and therapeutics. Furthermore, I propose to develop additional sophisticated graph theoretic techniques to mine network data and hence complement biological information that can be extracted from sequence. I propose to exploit these new techniques for biological applications in collaboration with experimentalists at Imperial College London: 1. aligning biological networks of species whose genomes are closely related, but that have very different phenotypes, in order to uncover systems-level factors that contribute to pronounced differences; 2. compare and contrast stress response pathways and metabolic pathways in bacteria in a unified systems-level framework and exploit the findings for: (a) bioengineering of micro-organisms for industrial applications (production of bio-fuels, bioremediation, production of biopolymers); (b) biomedical applications.

1 638 175 €

Start date: 2012-01-01, End date: 2017-12-31