Scientific Publications
Detach-ROCKET: Sequential feature selection for time series classification with random convolutional kernels
Abstract: Time Series Classification (TSC) is essential in fields like medicine, environmental science, and finance, enabling tasks such as disease diagnosis, anomaly detection, and stock price analysis. While machine learning models like Recurrent Neural Networks and InceptionTime are successful in numerous applications, they can face scalability issues due to computational requirements. Recently, ROCKET has emerged as an efficient alternative, achieving state-of-the-art performance and simplifying training by utilizing a large number of randomly generated features from the time series data. However, many of these features are redundant or non-informative, increasing computational load and compromising generalization. Here we introduce Sequential Feature Detachment (SFD) to identify and prune non-essential features in ROCKET-based models, such as ROCKET, MiniRocket, and MultiRocket. SFD estimates feature importance using model coefficients and can handle large feature sets without complex hyperparameter tuning. Testing on the UCR archive shows that SFD can produce models with better test accuracy using only 10% of the original features. We named these pruned models Detach-ROCKET. We also present an end-to-end procedure for determining an optimal balance between the number of features and model accuracy. On the largest binary UCR dataset, Detach-ROCKET improves test accuracy by 0.6% while reducing features by 98.9%. By enabling a significant reduction in model size without sacrificing accuracy, our methodology improves computational efficiency and contributes to model interpretability. We believe that Detach-ROCKET will be a valuable tool for researchers and practitioners working with time series data, who can find a user-friendly implementation of the model at https://github.com/gon-uri/detach_rocket. Authors: Gonzalo Uribarri, Federico Barone, Alessio Ansuini, Eric Fransén Journal: Data Mining and Knowledge Discovery Publication date: 20/08/2024 Consult the publication
Probing conformational dynamics of EGFR mutants via SEIRA spectroscopy: potential implications for tyrosine kinase inhibitor design
Abstract Missense mutations in EGFR’s catalytic domain alter its function, promoting cancer. SEIRA spectroscopy, supported by MD simulations, reveals structural differences in the compactness and hydration of helical motifs between active and inactive EGFR conformations models. These findings provide novel insights into the biophysical mechanisms driving EGFR activation and drug resistance, offering a robust method for studying emerging EGFR mutations and their structural impacts on TKIs efficacy. Authors Emiliano Laudadio, Federica Piccirilli, Henrick Vondracek, Giovanna Mobbili, Marta Stefania Semrau, Paola Storici, Roberta Galeazzi, Elena Romagnoli, Leonardo Sorci, Andrea Toma, Vincenzo Aglieri, Giovanni Birarda, Cristina Minnelli Journal Physical Chemistry Chemical Physics (PCCP) Publication Date 19/08/2024 Consult the publication
La0.2Sr0.25Ca0.45TiO3 Surface Reactivity with H2: A Combined Operando NEXAFS and Computational Study
Abstract A-site doped SrTiO3 is considered as a promising substitute for traditional anodic metals in solid oxide fuel cells (SOFCs). In this study, we present the reactivity of La0.2Sr0.25Ca0.45TiO3 (LCSTO), La0.2Sr0.7TiO3 (LSTO), and SrTiO3 (STO) toward H2 by operando ambient pressure NEXAFS spectroscopy and theoretical spectra simulation with FDMNES code. The samples were synthesized by MBE (molecular beam epitaxy), hydrothermal, and modified-Pechini routes. We found that the reducibility of the samples depends not only on their stoichiometry but also on the morphology, which is determined by the synthetic method. The results of these experiments give insight into the reducibility of Ti4+ in perovskites as well as the opportunity to further optimize the synthesis of these materials to obtain the best performance for SOFC applications. Authors F. Bassato, S. Mauri, L. Braglia, A. Yu. Petrov, E. Dobovičnik, F. Tavani, A. Tofoni, P. Ferrer, D. Grinter, G. Held, P. D’Angelo, P. Torelli Journal The Journal of Physical Chemistry Letters Date 13/08/2024 Consult the paper
Molecular findings and virological assessment of bladder papillomavirus infection in cattle.
Abstract Bovine and ovine papillomaviruses (BPVs – OaPVs) are infectious agents that have an important role in bladder carcinogenesis of cattle. In an attempt to better understand territorial prevalence of papillomavirus genotypes and gain insights into their molecular pathway(s), a virological assessment of papillomavirus infection was performed on 52 bladder tumors in cattle using droplet digital polymerase chain reaction (ddPCR), an improved version of conventional PCR. ddPCR detected and quantified BPV DNA and mRNAs in all tumor samples, showing that these viruses play a determinant role in bovine bladder carcinogenesis. OaPV DNA and mRNA were detected and quantified in 45 bladder tumors. BPV14, BPV13, BPV2, OaPV2, OaPV1, and OaPV3 were the genotypes most closely related to bladder tumors. ddPCR quantified BPV1 and OaPV4 DNA and their transcripts less frequently. Western blot analysis revealed a significant overexpression of the phosphorylated platelet derived growth factor β receptor (PDGFβR) as well as the transcription factor E2F3, which modulate cell cycle progression in urothelial neoplasia. Furthermore, significant overexpression of calpain1, a Cys protease, was observed in bladder tumors related to BPVs alone and in BPV and OaPV coinfection. Calpain1 has been shown to play a role in producing free transcription factors of the E2F family, and molecular findings suggest that calpain family members work cooperatively to mutually regulate their protease activities in cattle bladder tumors. Altogether, these results showed territorial prevalence of BPV and OaPV genotypes and suggested that PDGFβR and the calpain system appeared to be molecular partners of both BPVs and OaPVs. Authors Francesca De Falco, Anna Cutarelli, Francesca Luisa Fedele, Cornel Catoi, Sante Roperto Journal Veterinary Quarterly Publication date 04/08/2024 Consult the publication
Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals
Abstract Interpretability research aims to bridge the gap between empirical success and our scientific understanding of the inner workings of large language models (LLMs). However, most existing research focuses on analyzing a single mechanism, such as how models copy or recall factual knowledge. In this work, we propose a formulation of competition of mechanisms, which focuses on the interplay of multiple mechanisms instead of individual mechanisms and traces how one of them becomes dominant in the final prediction. We uncover how and where mechanisms compete within LLMs using two interpretability methods: logit inspection and attention modification. Our findings show traces of the mechanisms and their competition across various model components and reveal attention positions that effectively control the strength of certain mechanisms. Authors Francesco Ortu, Zhijing Jin, Diego Doimo, Mrinmaya Schaan, Alberto Cazzaniga, Bernhard Scholkopf Journal Accepted at the Annual Meeting of the Association for Computational Linquistics (ACL), Arxiv preprint: 2402.11655 Date 06/06/2024 Consult the paper
Emergent representations in networks trained with the Forward-Forward algorithm
Abstract The Backpropagation algorithm has often been criticised for its lack of biological realism. In an attempt to find a more biologically plausible alternative, the recently introduced Forward-Forward algorithm replaces the forward and backward passes of Backpropagation with two forward passes. In this work, we show that the internal representations obtained by the Forward-Forward algorithm can organise into category-specific ensembles exhibiting high sparsity – composed of a low number of active units. This situation is reminiscent of what has been observed in cortical sensory areas, where neuronal ensembles are suggested to serve as the functional building blocks for perception and action. Interestingly, while this sparse pattern does not typically arise in models trained with standard Backpropagation, it can emerge in networks trained with Backpropagation on the same objective proposed for the Forward-Forward algorithm. These results suggest that the learning procedure proposed by Forward-Forward may be superior to Backpropagation in modelling learning in the cortex, even when a backward pass is used. Authors Niccolò Tosato, Lorenzo Basile, Emanuele Ballarin, Giuseppe de Alteriis, Alberto Cazzaniga, Alessio Ansuini Journal Submitted at Advances in Neural Information Processing Systems 37 (NEURIPS 2024) Main Conference Track. Arxiv preprint: 2305.18353 Date 19/06/2024 Consult the paper
Enhancing Multi-Tip Artifact Detection in STM Images Using Fourier Transform and Vision Transformers
Abstract We address the issue of multi-tip artifacts in Scanning Tunneling Microscopy (STM) images by applying the fast Fourier transform (FFT) as a feature engineering method. We fine-tune various neural network architectures using a synthetic dataset, including Vision Transformers (ViT). The FFT-based preprocessing significantly improves the performance of ViT models compared to using only the grayscale channel. Ablation experiments highlight the optimal conditions for synthetic dataset generation. Unlike traditional methods that are challenging to implement for large datasets and used offline, our method enables on-the-fly classification at scale. Our findings demonstrate the efficacy of combining the Fourier transform with deep learning for enhanced artifact detection in STM images, contributing to more accurate analysis in material science research. Authors Tommaso Rodani, Alessio Ansuini, Alberto Cazzaniga Journal ICML ’24 Workshop ML for Life and Material Science: From Theory to Industry Applications Publication date 17/07/2024 Consult the paper
Enhancing predictions of protein stability changes induced by single mutations using MSA-based language models
Abstract Protein language models offer a new perspective for addressing challenges in structural biology, while relying solely on sequence information. Recent studies have investigated their effectiveness in forecasting shifts in thermodynamic stability caused by single amino acid mutations, a task known for its complexity due to the sparse availability of data, constrained by experimental limitations. To tackle this problem, we introduce two key novelties: leveraging a protein language model that incorporates Multiple Sequence Alignments to capture evolutionary information, and using a recently released mega-scale dataset with rigorous data preprocessing to mitigate overfitting. Authors Francesca Cuturello, Marco Celoria, Alessio Ansuini, Alberto Cazzaniga Journal Bioinformatics, 2024, 40 (7) Consult the paper
Operando Soft X-ray Absorption of LaMn1–xCoxO3 Perovskites for CO Oxidation
Abstract We employed operando soft X-ray absorption spectroscopy (XAS) to monitor the changes in the valence states and spin properties of LaMn1–xCoxO3 catalysts subjected to a mixture of CO and O2 at ambient pressure. Guided by simulations based on charge transfer multiplet theory, we quantitatively analyze the Mn and Co 2p XAS as well as the oxygen K-edge XAS spectra during the reaction process. The Mn sites are particularly sensitive to the catalytic reaction, displaying dynamics in their oxidation state. When Co doping is introduced (x ≤ 0.5), Mn oxidizes from Mn2+ to Mn3+ and Mn4+, while Co largely maintains a valence state of Co2+. In the case of LaCoO3, we identify high-spin and low-spin Co3+ species combined with Co2+. Our investigation underscores the importance to consider the spin and valence states of catalyst materials under operando conditions. Authors Qijun Che, Mahnaz Ghiasi, Luca Braglia, Matt LJ Peerlings, Silvia Mauri, Piero Torelli, Petra de Jongh, Frank MF de Groot Journal ACS Catalysis Date 12/07/2024 Consult the paper
Molecular detection of transcriptionally active ovine papillomaviruses in commercial equine semen
Abstract Virological evaluation was performed on equine semen to detect the presence of papillomaviruses (PVs) using droplet digital polymerase chain reaction (ddPCR) as the aim of this study was to investigate whether the sperm from asymptomatic stallions harbors ovine papillomaviruses (OaPVs). Twenty-seven semen samples were analyzed, 18 of which were commercially acquired. The remaining nine samples comprising semen and peripheral blood, were collected from nine stallions with no apparent signs of PV-related diseases during clinical examination at the Didactic Veterinary University Hospital (DVUH) of Naples. OaPV was detected in 26 semen samples. OaPV1 was the most prevalent virus infecting equine semen. OaPV1 infected 21 semen samples (~80.8%) and showed a high number of DNA and RNA copies per microliter. qPCR was used to detect OaPV1 DNA in the 18 semen samples. ddPCR was used to detect and quantify the expression of OaPV2, OaPV3, and OaPV4. qPCR failed to detect DNA for these genotypes. Additionally, ddPCR was used to detect the transcriptionally active OaPV1 in six blood and semen samples from the same stallion. ddPCR failed to detect any nucleic acids in OaPVs in peripheral blood samples from the three stallions. In one semen sample, ddPCR detected OaPV1 DNA but failed to detect any nucleic acid in the remaining two semen samples, and peripheral blood from the same animals of the remaining 18 semen samples was not available, OaPV1 and OaPV4 were responsible for nine and five single infections, respectively. No single infections with either OaPV3 or OaPV4 were seen. Authors Anna Cutarelli, Francesca De Falco, Roberta Brunetti, Michele Napoletano, Giovanna Fusco, Sante Roperto Journal Frontiers in Veterinary Science Publication Date 03/07/2024 Consult the publication
Genome-wide DNA methylation and transcriptomic analysis of liver tissues subjected to early ischemia/reperfusion injury upon human liver transplantation
Abstract Introduction and Objectives Epigenetic changes represent a mechanism connecting external stresses with long-term modifications of gene expression programs. In solid organ transplantation, ischemia-reperfusion injury (IRI) appears to induce epigenomic changes in the graft, although the currently available data are extremely limited. The present study aimed to characterize variations in DNA methylation and their effects on the transcriptome in liver transplantation from brain-dead donors. Patients and Methods 12 liver grafts were evaluated through serial biopsies at different timings in the procurement-transplantation process: T0 (warm procurement, in donor), T1 (bench surgery), and T2 (after reperfusion, in recipient). DNA methylation (DNAm) and transcriptome profiles of biopsies were analyzed using microarrays and RNAseq. Results Significant variations in DNAm were identified, particularly between T2 and T0. Functional enrichment of the best 1000 ranked differentially methylated promoters demonstrated that 387 hypermethylated and 613 hypomethylated promoters were involved in spliceosomal assembly and response to biotic stimuli, and inflammatory immune responses, respectively. At the transcriptome level, T2 vs. T0 showed an upregulation of 337 and downregulation of 61 genes, collectively involved in TNF-α, NFKB, and interleukin signaling. Cell enrichment analysis individuates macrophages, monocytes, and neutrophils as the most significant tissue-cell type in the response. Conclusions In the process of liver graft procurement-transplantation, IRI induces significant epigenetic changes that primarily act on the signaling pathways of inflammatory responses dependent on TNF-α, NFKB, and interleukins. Our DNAm datasets are the early IRI methylome literature and will serve as a launch point for studying the impact of epigenetic modification in IRI. Authors Giraudi PJ, Laraño AA, Monego SD, Pravisani R, Bonazza D, Gondolesi G, Tiribelli C, Baralle F, Baccarani U, Licastro D. Journal Ann Hepatol. Date 01/07/2024 Consult the publication
Low-Temperature Methane Activation Reaction Pathways over Mechanochemically-Generated Ce4+/Cu+ Interfacial Sites
Abstract Methane is a valuable resource and its valorization is an important challenge in heterogeneous catalysis. Here it is shown that CeO2/CuO composite prepared by ball milling activates methane at a temperature as low as 250 °C. In contrast to conventionally prepared catalysts, the formation of partial oxidation products such as methanol and formaldehyde is also observed. Through an in situ Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS) and operando Near Edge X-Ray Absorption Fine Structure Spectroscopy (NEXAFS) approach, it can be established that this unusual reactivity can be attributed to the presence of Ce4+/Cu+ interfaces generated through a redox exchange between Ce3+ and Cu2+ atoms facilitated by the mechanical energy supplied during milling. DFT modeling of the electronic properties confirms the existence of a charge transfer mechanism. These results demonstrate the effectiveness and distinctiveness of the mechanical approach in creating unique and resilient interfaces thereby enabling the optimization and refining of CeO2/CuO catalysts in methane activation reactions. Authors Silvia Mauri, Rudy Calligaro, Carlo Federico Pauletti, Matteo Farnesi Camellone, Marta Boaro, Luca Braglia, Stefano Fabris, Simone Piccinin, Piero Torelli, Alessandro Trovarelli Journal Small Date 11/06/2024 Consult the paper