Search

Molecular docking is a widely used technique in drug discovery to predict the binding mode of a given ligand to its target. However, the identification of the near-native binding pose in docking experiments still represents a challenging task as the scoring functions currently employed by docking programs are parametrized to predict the binding affinity, and, therefore, they often fail to correctly identify the ligand native binding conformation. Selecting the correct binding mode is crucial to obtaining meaningful results and to conveniently optimizing new hit compounds. Deep learning (DL) algorithms have been an area of a growing interest in this sense for their capability to extract the relevant information directly from the protein-ligand structure. Our review aims to present the recent advances regarding the development of DL-based pose selection approaches, discussing limitations and possible future directions. Moreover, a comparison between the performances of some classical scoring functions and DL-based methods concerning their ability to select the correct binding mode is reported. In this regard, two novel DL-based pose selectors developed by us are presented.

The widespread use of visual surveillance in public areas puts individual privacy at stake while also increasing resource usage (energy, bandwidth, and computation). Neuromorphic vision sensors (or event cameras) are considered viable solutions for privacy issues; since event cameras only capture scene dynamics, they do not capture detailed RGB images of individuals. However, recent deep learning architectures have enabled the reconstruction of high-fidelity images from event sensor data that could reveal individual identity information. As a result, it reintroduces privacy risks for event-based vision applications. In this work, we focus on protecting the identity of individuals from such image reconstruction attacks by anonymizing event data. To achieve this, we present an end-to-end network architecture jointly optimized for the twofold objective of preserving privacy and performing a downstream computer vision task. The proposed network learns to scramble events, thereby degrading the quality of images that potential intruders could reconstruct. We demonstrate the application of our framework in two challenging computer vision tasks: person re-identification (ReId) and human pose estimation (HPE). To this end, we also present and evaluate the first event-based person ReId dataset, Event-ReId. We validate the privacy-preserving efficacy of our approach against possible privacy attacks through extensive experiments: for person ReId, we utilize the real event-based Event-ReId dataset and synthetic event data simulated from the SoftBio dataset; for HPE, we use a publicly available event-based dataset DHP19. The results of both tasks show that anonymizing event data effectively protects private information with minimal impact on the subsequent task performance.

Active objects are those in contact with the first person in an egocentric video. This paper addresses the challenge of anticipating the future location of the next active object in relation to a person within a given egocentric video clip, which is challenging since the contact is poised to happen after the last observed frame by the model, even before any action takes place. As we aim to estimate the position of objects, this problem is particularly hard in a scenario where the observed clip and the action segment are separated by the so-called time-to-contact segment. We term this task Anticipating the Next ACTive Object (ANACTO) and introduce a transformer-based self-attention framework to tackle it. We compare our model with the existing anticipation-based methods to establish relevant baseline methods, where our approach outperforms all of them on three major egocentric datasets: EpicKitchens-100, EGTEA+, and Ego4D. We also conduct an ablation study to better present the effectiveness of the proposed and baseline methods on varying conditions. The code as well as the ANACTO task annotations for the aforementioned first two datasets will be made available upon the acceptance of this paper.

Abstract In this paper, we investigate brain activity associated with complex visual tasks, showing that electroencephalography (EEG) data can help computer vision in reliably recognizing actions from video footage that is used to stimulate human observers. Notably, we consider not only typical “explicit” video action benchmarks, but also more complex data sequences in which action concepts are only referred to, implicitly. To this end, we consider a challenging action recognition benchmark dataset—Moments in Time—whose video sequences do not explicitly visualize actions, but only implicitly refer to them (e.g., fireworks in the sky as an extreme example of “flying”). We employ such videos as stimuli and involve a large sample of subjects to collect a high-definition, multi-modal EEG and video data, designed for understanding action concepts. We discover an agreement among brain activities of different subjects stimulated by the same video footage. We name it as subjects consensus, and we design a computational pipeline to transfer knowledge from EEG to video, sharply boosting the recognition performance.

Our brain constantly combines sensory information in unitary percept to build coherent representations of the environment. Even though this process could appear smooth, integrating sensory inputs from various sensory modalities must overcome several computational issues, such as recoding and statistical inferences problems. Following these assumptions, we developed a neural architecture replicating humans' ability to use audiovisual spatial representations. We considered the well-known ventriloquist illusion as a benchmark to evaluate its phenomenological plausibility. Our model closely replicated human perceptual behavior, proving a truthful approximation of the brain's ability to develop audiovisual spatial representations. Considering its ability to model audiovisual performance in a spatial localization task, we release our model in conjunction with the dataset we recorded for its validation. We believe it will be a powerful tool to model and better understand multisensory integration processes in experimental and rehabilitation environments.

Fermentation is a widely used process in the biotechnology industry, in which sugar-based substrates are transformed into a new product through chemical reactions carried out by microorganisms. Fermentation yields depend heavily on critical process parameter (CPP) values which need to be finely tuned throughout the process; this is usually performed by a biotech production expert relying on empirical rules and personal experience. Although developing a mathematical model to analytically describe how yields depend on CPP values is too challenging because the process involves living organisms, we demonstrate the benefits that can be reaped by using a black-box machine learning (ML) approach based on recurrent neural networks (RNN) and long short-term memory (LSTM) neural networks to predict real time OD600nm values from fermentation CPP time series. We tested both networks on an E. coli fermentation process (upstream) optimized to obtain inclusion bodies whose purification (downstream) in a later stage will yield a targeted neurotrophin recombinant protein. We achieved root mean squared error (RMSE) and relative error on final yield (REFY) performances which demonstrate that RNN and LSTM are indeed promising approaches for real-time, in-line process yield estimation, paving the way for machine learning-based fermentation process control algorithms.