Search

Explainable artificial intelligence has become a vitally important research field aiming, among other tasks, to justify predictions made by intelligent classifiers automatically learned from data. Importantly, efficiency of automated explanations may be undermined if the end user does not have sufficient domain knowledge or lacks information about the data used for training. To address the issue of effective explanation communication, we propose a novel information-seeking explanatory dialogue game following the most recent requirements to automatically generated explanations. Further, we generalise our dialogue model in form of an explanatory dialogue grammar which makes it applicable to interpretable rule-based classifiers that are enhanced with the capability to provide textual explanations. Finally, we carry out an exploratory user study to validate the corresponding dialogue protocol and analyse the experimental results using insights from process mining and argument analytics. A high number of requests for alternative explanations testifies the need for ensuring diversity in the context of automated explanations.

Auxin is a crucial growth regulator that governs plant development and responses to environmental perturbations. It functions at the heart of many developmental processes, from embryogenesis to organ senescence, and is key to plant interactions with the environment, including responses to biotic and abiotic stimuli. As remarkable as auxin is, it does not act alone, but rather solicits the help of, or is solicited by, other endogenous signals, including the plant hormones abscisic acid, brassinosteroids, cytokinins, ethylene, gibberellic acid, jasmonates, salicylic acid, and strigolactones. The interactions between auxin and other hormones occur at multiple levels: hormones regulate one another's synthesis, transport, and/or response; hormone-specific transcriptional regulators for different pathways physically interact and/or converge on common target genes; etc. However, our understanding of this crosstalk is still fragmentary, with only a few pieces of the gigantic puzzle firmly established. In this review, we provide a glimpse into the complexity of hormone interactions that involve auxin, underscoring how patchy our current understanding is.

The explanatory capacity of interpretable fuzzy rule-based classifiers is usually limited to offering explanations for the predicted class only. A lack of potentially useful explanations for non-predicted alternatives can be overcome by designing methods for the so-called counterfactual reasoning. Nevertheless, state-of-the-art methods for counterfactual explanation generation require special attention to human evaluation aspects, as the final decision upon the classification under consideration is left for the end user. In this paper, we first introduce novel methods for qualitative and quantitative counterfactual explanation generation. Then, we carry out a comparative analysis of qualitative explanation generation methods operating on (combinations of) linguistic terms as well as a quantitative method suggesting precise changes in feature values. Then, we propose a new metric for assessing the perceived complexity of the generated explanations. Further, we design and carry out two human evaluation experiments to assess the explanatory power of the aforementioned methods. As a major result, we show that the estimated explanation complexity correlates well with the informativeness, relevance, and readability of explanations perceived by the targeted study participants. This fact opens the door to using the new automatic complexity metric for guiding multi-objective evolutionary explainable fuzzy modeling in the near future Ilia Stepin is an FPI researcher (grant PRE2019-090153). Jose M. Alonso-Moral is a Ramon y Cajal researcher (grant RYC-2016–19802). This work was supported by the Spanish Ministry of Science and Innovation (grants RTI2018-099646-B-I00, PID2021-123152OB-C21, and TED2021-130295B-C33) and the Galician Ministry of Culture, Education, Professional Training and University (grants ED431F2018/02, ED431G2019/04, and ED431C2022/19). All the grants were co-funded by the European Regional Development Fund (ERDF/FEDER program). SI

This paper presents the art painting style explainable classifier named ANYXI. The classifier is based on art specialists’ knowledge of art styles and human-understandable color traits. ANYXI overcomes the principal flaws in the few art painting style classifiers in the literature. In this way, we first propose, using the art specialists’ studies, categorizations of the Baroque, Impressionism, and Post-Impressionism. Second, we carry out a human survey with the aim of validating the appropriateness of the color features used in the categorizations for human understanding. Then, we analyze and discuss the accuracy and interpretability of the ANYXI classifier. The study ends with an evaluation of the rationality of explanations automatically generated by ANYXI. We enrich the discussion and empirical validation of ANYXI by considering a quantitative and qualitative comparison versus other explainable classifiers. The reported results show how ANYXI is outstanding from the point of view of interpretability while keeping high accuracy (comparable to non-explainable classifiers). Moreover, automated generations are endowed with a good level of rationality.

Plants often live in adverse environmental conditions and are exposed to various stresses, such as heat, cold, heavy metals, salt, radiation, poor lighting, nutrient deficiency, drought, or flooding. To adapt to unfavorable environments, plants have evolved specialized molecular mechanisms that serve to balance the trade-off between abiotic stress responses and growth. These mechanisms enable plants to continue to develop and reproduce even under adverse conditions. Ethylene, as a key growth regulator, is leveraged by plants to mitigate the negative effects of some of these stresses on plant development and growth. By cooperating with other hormones, such as jasmonic acid (JA), abscisic acid (ABA), brassinosteroids (BR), auxin, gibberellic acid (GA), salicylic acid (SA), and cytokinin (CK), ethylene triggers defense and survival mechanisms thereby coordinating plant growth and development in response to abiotic stresses. This review describes the crosstalk between ethylene and other plant hormones in tipping the balance between plant growth and abiotic stress responses.

In this work, we present a complete system to produce an automatic linguistic reporting about the customer activity patterns inside open malls, a mixed distribution of classical malls joined with the shops on the street. These reports can assist to design marketing campaigns by means of identifying the best places to catch the attention of customers. Activity patterns are estimated with process mining techniques and the key information of localization. Localization is obtained with a parallelized solution based on WiFi fingerprint system to speed up the solution. In agreement with the best practices for human evaluation of natural language generation systems, the linguistic quality of the generated report was evaluated by 41 experts who filled in an online questionnaire. Results are encouraging, since the average global score of the linguistic quality dimension is 6.17 (0.76 of standard deviation) in a 7- point Likert scale. This expresses a high degree of satisfaction of the generated reports and validates the adequacy of automatic natural language textual reports as a complementary tool to process model visualization This work has been partially supported by the Spanish Ministry of Science Innovation and Universities and the European Regional Development Fund (ERDF/FEDER) Grants RTI2018-099646-BI00, TIN2017-84796-C2-1-R, TIN2017-90773-REDT, RED2018-102641-T and RYC-2016-19802 (Ramón y Cajal program, José M. Alonso). Also by the Galician Ministry of Education, University and Professional Training and the ERDF/FEDER program (ED431F2018/02, ED431C2018/29, ED431G2019/04 grants) SI

BACKGROUND: High intraoperative PEEP with recruitment manoeuvres may improve perioperative outcomes. We re-examined this question by conducting a patient-level meta-analysis of three clinical trials in adult patients at increased risk for postoperative pulmonary complications who underwent non-cardiothoracic and non-neurological surgery. METHODS: The three trials enrolled patients at 128 hospitals in 24 countries from February 2011 to February 2018. All patients received volume-controlled ventilation with low tidal volume. Analyses were performed using one-stage, two-level, mixed modelling (site as a random effect; trial as a fixed effect). The primary outcome was a composite of postoperative pulmonary complications within the first week, analysed using mixed-effect logistic regression. Pre-specified subgroup analyses of nine patient characteristics and seven procedure and care-delivery characteristics were also performed. RESULTS: Complete datasets were available for 1913 participants ventilated with high PEEP and recruitment manoeuvres, compared with 1924 participants who received low PEEP. The primary outcome occurred in 562/1913 (29.4%) participants randomised to high PEEP, compared with 620/1924 (32.2%) participants randomised to low PEEP (unadjusted odds ratio [OR]=0.87; 95% confidence interval [95% CI], 0.75-1.01; P=0.06). Higher PEEP resulted in 87/1913 (4.5%) participants requiring interventions for desaturation, compared with 216/1924 (11.2%) participants randomised to low PEEP (OR=0.34; 95% CI, 0.26-0.45). Intraoperative hypotension was associated more frequently (784/1913 [41.0%]) with high PEEP, compared with low PEEP (579/1924 [30.1%]; OR=1.87; 95% CI, 1.60-2.17). CONCLUSIONS: High PEEP combined with recruitment manoeuvres during low tidal volume ventilation in patients undergoing major surgery did not reduce postoperative pulmonary complications. CLINICAL TRIAL REGISTRATION: NCT03937375 (Clinicaltrials.gov).

LaMnO3+δ is one of the classical-formal nonstoichiometric systems in materials science that, for δ = 0, shows antiferromagnetic ordering by super-exchange interactions between Mn3+ ions. Nevertheless, the chemical reactivity of solids goes through different pathways when the particle size is reduced to the nanometric range, leading to the modification of the physico-chemical properties of materials. It is demonstrated here that octagonal LaMnO3+δ nanoparticles of average size 20 nm significantly modify their non-stoichiometry, microstructure and magnetotransport properties and are thus able to tune their behavior, as a function of the cationic vacancy concentration. For δ = 0, LaMnO3 nanoparticles, even with only Mn3+, become ferromagnetic, whereas for δ = 0.23, i.e. La0.93Mn0.93O3, due to a cationic diffusion mechanism, weaker ferromagnetic interactions appear promoting the emergence of magnetoresistance. This accommodation of compositional variations triggers Mn3+ → Mn4+ oxidation and facilitates, as observed by atomically resolved scanning transmission electron microscopy and electron energy loss spectroscopy, the displacement of some La atoms around their normal site that can introduce restrictions in double exchange Mn3+–O2−–Mn4+ interactions.

A number of algorithms in the field of artificial intelligence offer poorly interpretable decisions. To disclose the reasoning behind such algorithms, their output can be explained by means of so-called evidence-based (or factual) explanations. Alternatively, contrastive and counterfactual explanations justify why the output of the algorithms is not any different and how it could be changed, respectively. It is of crucial importance to bridge the gap between theoretical approaches to contrastive and counterfactual explanation and the corresponding computational frameworks. In this work we conduct a systematic literature review which provides readers with a thorough and reproducible analysis of the interdisciplinary research field under study. We first examine theoretical foundations of contrastive and counterfactual accounts of explanation. Then, we report the state-of-the-art computational frameworks for contrastive and counterfactual explanation generation. In addition, we analyze how grounded such frameworks are on the insights from the inspected theoretical approaches. As a result, we highlight a variety of properties of the approaches under study and reveal a number of shortcomings thereof. Moreover, we define a taxonomy regarding both theoretical and practical approaches to contrastive and counterfactual explanation.

Protein production from mRNA is one of the fundamental molecular processes in a cell. Accurate genome-wide information on the levels of translation and ribosome distribution on mRNA can be gathered by carrying out ribosome footprinting, aka Ribo-seq. Herein, we present a detailed protocol describing the construction of parallel Ribo-seq and RNA-seq libraries from Arabidopsis seedlings treated with the plant hormone auxin. The improved protocol for ribosome footprint library generation can be easily adapted to analyzing the effects on translation of genetic perturbations and various abiotic and biotic factors to shed the much-needed light on translational regulation in plants.

Multicellular organisms use dedicator of cytokinesis (DOCK) family guanine nucleotide exchange factors (GEFs) to activate Rac/Rho-of-plants small GTPases and coordinate cell shape change. In developing tissues, DOCK signals integrate cell-cell interactions with cytoskeleton remodeling, and the GEFs cluster reversibly at specific organelle surfaces to orchestrate cytoskeletal reorganization. The domain organizations among DOCK orthologs are diverse, and the mechanisms of localization control are poorly understood. Here, we use combinations of transgene complementation and live-cell imaging assays to uncover an evolutionarily conserved and essential localization determinant in the DOCK-GEF named SPIKE1. The SPIKE1-DHR3 domain is sufficient for organelle association in vivo, and displays a complicated lipid-binding selectivity for both phospholipid head groups and fatty acid chain saturation. SPIKE1-DHR3 is predicted to adopt a C2-domain structure and functions as part of a tandem C2 array that enables reversible clustering at the cell apex. This work provides mechanistic insight into how DOCK GEFs sense compositional and biophysical membrane properties at the interface of two organelle systems.

[EN] This paper presents a new series expansion based on Bernoulli matrix polynomials to approximate the matrix cosine function. An approximation based on this series is not a straightforward exercise since there exist different options to implement such a solution. We dive into these options and include a thorough comparative of performance and accuracy in the experimental results section that shows benefits and downsides of each one. Also, a comparison with the Pade approximation is included. The algorithms have been implemented in MATLAB and in CUDA for NVIDIA GPUs., Spanish Ministerio de Economia y Competitividad and European Regional Development Fund, Grant/Award Number: TIN2017-89314-P; Universitat Politecnica de Valencia, Grant/Award Number: SP20180016

The allene functional group and the indole ring have emerged as recurring building blocks in organic chemistry due to their diverse reactivity and straightforward preparation. In addition, the biological properties and natural occurrence of indole give a remarkable added-value to newly-prepared indole-based structures. In the last decade, the combination of allenes and indoles as co-reactants has revealed a fruitful reactivity allowing the synthesis of complex structures and the investigation of new mechanistic pathways, such as the challenging asymmetric nucleophilic addition using optically pure allenes as chirality transfer agents, the double nucleophilic addition through metal carbene intermediates, or novel hybrid catalysis in a bimetallic system. Herein we report a critical review into the wide and fascinating ways allenes and indoles can interact, and the structural diversity that can be therefore achieved.

The complex nature of multifactorial diseases, such as Morbus Alzheimer, has produced a strong need to design multitarget-directed ligands to address the involved complementary pathways. We performed a purposive structural modification of a tetratarget small-molecule, that is contilisant, and generated a combinatorial library of 28 substituted chromen-4-ones. The compounds comprise a basic moiety which is linker-connected to the 6-position of the heterocyclic chromenone core. The syntheses were accomplished by Mitsunobu- or Williamson-type ether formations. The resulting library members were evaluated at a panel of seven human enzymes, all of which being involved in the pathophysiology of neurodegeneration. A concomitant inhibition of human acetylcholinesterase and human monoamine oxidase B, with IC50 values of 5.58 and 7.20 μM, respectively, was achieved with the dual-target 6-(4-(piperidin-1-yl)butoxy)-4H-chromen-4-one (7).

[EN] Matrix differential equations are at the heart of many science and engineering problems. In this paper, a procedure based on higher-order matrix splines is proposed to provide the approximated numerical solution of special nonlinear third-order matrix differential equations, having the form Y-(3)(x)=f(x,Y(x)). Some numerical test problems are also included, whose solutions are computed by our method., This research was partially funded by the European Regional Development Fund (ERDF) and the Spanish Ministerio de Economia y Competitividad Grant TIN2017-89314-P

DiSSCo (Distributed System of Scientific Collections) is a research infrastructure (RI) under development, which will provide services for the global research community to support and enhance physical and digital access to the natural history collections in Europe. These services include training, support, documentation and e-services. This talk will focus on the e-services and will give an overview of the current status, roadmap and first results as an introduction to the next talks in the session, which focus on some of the services in more detail and the standards work undertaken in Biodiversity Information Standards (TDWG) to enable them. The RI community will provide the envisioned e-services, which will use the novel FAIR Digital Object (FDO) infrastructure serving digital specimens from the European collections. The infrastructure will provide integrated data analysis, enhanced interpretation, annotation and access services for community curation and visualisation. The FDO infrastructure enables specimen data to be (re-)connected with genomic, geographical, morphological, taxonomic and environmental information through the digital specimen, making them Digital Extended Specimens. A large number of user stories have been collected through the DiSSCo-linked projects ICEDIG, SYNTHESYS+ and DiSSCo Prepare, to guide which e-Services to build and what functionality to provide. These user stories are publicly available in a github repository. The e-services are developed based on the user stories and prioritization provided by collection providers and the scientific community. A variety of mechanisms are used to collect input: surveys, workshops, roundtables and workpackage meetings, and feedback from users that have already been using beta versions of some of the services. DiSSCo aims to become operational in 2026 but several of the services are already being piloted or implemented. Experimental services and demonstrators are publicly available through DiSSCo Labs for testing and feedback. By connecting the specimen data with derived and related information in a FAIR way (Findable, Accessible, Interoperable and Reusable), the e-services will accelerate biodiversity discovery and support novel research questions. The FDO infrastructure has a data model that also integrates the PROV Ontology (PROV-O), which allows for the e-services to capture activities and improve the visibility of researcher contributions. This vision towards FAIR and high quality data is essential for community curation of the specimen data and making better use of the limited number of experts available. To provide the DiSSCo e-services in a FAIR way, the data derived from the natural history collections in Europe needs to be integrated as one virtual collection. The data has to be findable and accessible as soon as it is being created for services like a Specimen Data Refinery prior to publication in a facility like GBIF (Global Biodiversity Information Facility). This requires new standards for describing collections and specimen data. Standards being created to fill these gaps are TDWG CD (Collection Descriptions) and TDWG MIDS (Minimum Information about a Digital Specimen). The DiSSCo e-Services vision brings the data, standards, and processes together to serve the user community.

Nucleophilic additions to allenes catalyzed by transition metals represent a powerful tool to obtain selective access to diverse structures with numerous applications. Reported here is a straightforward methodology to achieve selective addition of azoles to the proximal or distal carbon of activated allenes depending on the catalyst used, unravelling a gold-platinum bimetallic catalysis approach for the 1,3-double addition of azoles to activated allenes to give 1,3-bisazole derivatives, an important scaffold in medicinal and organometallic chemistry.

Ruddlesden–Popper oxides, (AO)(ABO3) n , occupy a prominent place in the landscape of materials research because of their intriguing potential applications. Compositional modifications to the cation sublattices, A or B, have been explored in order to achieve enhanced functionalities. However, changes to the anionic sublattice have been much less explored. In this work, new oxygen-deficient manganese Ruddlesden–Popper-related phases, La0.5Ca2.5Mn2O6.5 and La0.5Ca2.5Mn2O6.25, have been synthesized by controlled reduction of the fully oxidized n = 2 term La0.5Ca2.5Mn2O7. A complete structural and compositional characterization, by means of neutron diffraction, electron diffraction and atomically resolved scanning transmission electron microscopy and electron energy-loss spectroscopy techniques, allows the proposition of a topotactic reduction pathway through preferential oxygen removal in the [MnO2] layers along [031] and [0{\bar 1}3] directions. The gradual decrease of the Mn oxidation state, accommodated by short-range ordering of anionic vacancies, reasonably explains the breaking of ferromagnetic interactions reinforcing the emergence of antiferromagnetic ones. Additional short-range order–disorder phenomena of La and Ca cations have been detected in the reduced La0.5Ca2.5Mn2O7−δ, as previously reported in the parent compound.

Background Ribo-seq is a popular technique for studying translation and its regulation. A Ribo-seq experiment produces a snap-shot of the location and abundance of actively translating ribosomes within a cell’s transcriptome. In practice, Ribo-seq data analysis can be sensitive to quality issues such as read length variation, low read periodicities, and contaminations with ribosomal and transfer RNA. Various software tools for data preprocessing, quality assessment, analysis, and visualization of Ribo-seq data have been developed. However, many of these tools require considerable practical knowledge of software applications, and often multiple different tools have to be used in combination with each other. Results We present riboStreamR, a comprehensive Ribo-seq quality control (QC) platform in the form of an R Shiny web application. RiboStreamR provides visualization and analysis tools for various Ribo-seq QC metrics, including read length distribution, read periodicity, and translational efficiency. Our platform is focused on providing a user-friendly experience, and includes various options for graphical customization, report generation, and anomaly detection within Ribo-seq datasets. Conclusions RiboStreamR takes advantage of the vast resources provided by the R and Bioconductor environments, and utilizes the Shiny R package to ensure a high level of usability. Our goal is to develop a tool which facilitates in-depth quality assessment of Ribo-seq data by providing reference datasets and automatically highlighting quality issues and anomalies within datasets. Electronic supplementary material The online version of this article (10.1186/s12864-019-5700-7) contains supplementary material, which is available to authorized users.

Gene expression differences between distinct cell types are orchestrated by specific sets of transcription factors and epigenetic regulators acting upon the genome. In plants, the mechanisms underlying tissue-specific gene activity remain largely unexplored. Although transcriptional and epigenetic profiling of individual organs, tissues, and more recently, of single cells can easily detect the molecular signatures of different biological samples, how these unique cell identities are established at the mechanistic level is only beginning to be decoded. Computational methods, including machine learning, used in combination with experimental approaches, enable the identification and validation of candidate cis-regulatory elements driving cell-specific expression. Synthetic biology shows great promise not only as a means of testing candidate DNA motifs but also for establishing the general rules of nature driving promoter architecture and for the rational design of genetic circuits in research and agriculture to confer tissue-specific expression to genes or molecular pathways of interest.

On-surface synthesis has recently been regarded as a promising approach for the generation of new molecular structures. It has been particularly successful in the synthesis of graphene nanoribbons, nanographenes and intrinsically reactive and instable, yet attractive species. It is based on the combination of solution chemistry aimed at preparation of appropriate molecular precursors for further ultra-high vacuum surface assisted transformations. This approach also owes its success to an incredible development of characterization techniques, such as scanning tunneling/atomic force microscopy and related methods, which allow detailed, local characterization at atomic scale. While the surface-assisted synthesis can provide molecular nanostructures with outstanding precision, down to single atoms, it suffers from basing on metallic surfaces and often limited yield. Therefore, the extension of the approach away from metals and the struggle to increase productivity seem to be significant challenges toward wider applications. Herein, we demonstrate the on-surface synthesis approach for generation of non-planar nanographenes, which are synthesized through a combination of solution chemistry and sequential surface-assisted processes, together with the detailed characterization by scanning probe microscopy methods.

[EN] The most popular method for computing the matrix logarithm is a combination of the inverse scaling and squaring method in conjunction with a Pade approximation, sometimes accompanied by the Schur decomposition. In this work, we present a Taylor series algorithm, based on the free-transformation approach of the inverse scaling and squaring technique, that uses recent matrix polynomial formulas for evaluating the Taylor approximation of the matrix logarithm more efficiently than the Paterson-Stockmeyer method. Two MATLAB implementations of this algorithm, related to relative forward or backward error analysis, were developed and compared with different state-of-the art MATLAB functions. Numerical tests showed that the new implementations are generally more accurate than the previously available codes, with an intermediate execution time among all the codes in comparison., This research was funded by the European Regional Development Fund (ERDF) and the Spanish Ministerio de Economia y Competitividad Grant TIN2017-89314-P.

[EN] In this paper, we introduce two approaches to compute the matrix hyperbolic tangent. While one of them is based on its own definition and uses the matrix exponential, the other one is focused on the expansion of its Taylor series. For this second approximation, we analyse two different alternatives to evaluate the corresponding matrix polynomials. This resulted in three stable and accurate codes, which we implemented in MATLAB and numerically and computationally compared by means of a battery of tests composed of distinct state-of-the-art matrices. Our results show that the Taylor series-based methods were more accurate, although somewhat more computationally expensive, compared with the approach based on the exponential matrix. To avoid this drawback, we propose the use of a set of formulas that allows us to evaluate polynomials in a more efficient way compared with that of the traditional Paterson¿Stockmeyer method, thus, substantially reducing the number of matrix products (practically equal in number to the approach based on the matrix exponential), without penalising the accuracy of the result, This research was funded by the Spanish Ministerio de Ciencia e Innovacion under grant number TIN2017-89314-P.

The allene functionality has participated in one of the most exciting voyages in organic chemistry, from chemical curiosities to a recurring building block in modern organic chemistry. In the last decades, a special kind of allene, namely, allenol, has emerged. Allenols, formed by an allene moiety and a hydroxyl functional group with diverse connectivity, have become common building blocks for the synthesis of a wide range of structures and frequent motif in naturally occurring systems. The synergistic effect of the allene and hydroxyl functional groups enables allenols to be considered as a unique and sole functionality exhibiting a special reactivity. This Review summarizes the most significant contributions to the chemistry of allenols that appeared during the past decade, with emphasis on their synthesis, reactivity, and occurrence in natural products., This work was supported in part by AEI (MICIU) and FEDER (Project PGC2018-095025−B-I00). We thank Prof. Benito Alcaide and Prof. Pilar M. Fresneda for their mentorship in past years.

Alzheimer’s disease (AD) is the most common type of dementia. Its diagnosis and progression detection have been intensively studied. Nevertheless, research studies often have little effect on clinical practice mainly due to the following reasons: (1) Most studies depend mainly on a single modality, especially neuroimaging; (2) diagnosis and progression detection are usually studied separately as two independent problems; and (3) current studies concentrate mainly on optimizing the performance of complex machine learning models, while disregarding their explainability. As a result, physicians struggle to interpret these models, and feel it is hard to trust them. In this paper, we carefully develop an accurate and interpretable AD diagnosis and progression detection model. This model provides physicians with accurate decisions along with a set of explanations for every decision. Specifically, the model integrates 11 modalities of 1048 subjects from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) real-world dataset: 294 cognitively normal, 254 stable mild cognitive impairment (MCI), 232 progressive MCI, and 268 AD. It is actually a two-layer model with random forest (RF) as classifier algorithm. In the first layer, the model carries out a multi-class classification for the early diagnosis of AD patients. In the second layer, the model applies binary classification to detect possible MCI-to-AD progression within three years from a baseline diagnosis. The performance of the model is optimized with key markers selected from a large set of biological and clinical measures. Regarding explainability, we provide, for each layer, global and instance-based explanations of the RF classifier by using the SHapley Additive exPlanations (SHAP) feature attribution framework. In addition, we implement 22 explainers based on decision trees and fuzzy rule-based systems to provide complementary justifications for every RF decision in each layer. Furthermore, these explanations are represented in natural language form to help physicians understand the predictions. The designed model achieves a cross-validation accuracy of 93.95% and an F1-score of 93.94% in the first layer, while it achieves a cross-validation accuracy of 87.08% and an F1-Score of 87.09% in the second layer. The resulting system is not only accurate, but also trustworthy, accountable, and medically applicable, thanks to the provided explanations which are broadly consistent with each other and with the AD medical literature. The proposed system can help to enhance the clinical understanding of AD diagnosis and progression processes by providing detailed insights into the effect of different modalities on the disease risk This work was supported by National Research Foundation of Korea-Grant funded by the Korean Government (Ministry of Science and ICT)-NRF-2020R1A2B5B02002478). In addition, Dr. Jose M. Alonso is Ramon y Cajal Researcher (RYC-2016-19802), and its research is supported by the Spanish Ministry of Science, Innovation and Universities (grants RTI2018-099646-B-I00, TIN2017-84796-C2-1-R, TIN2017-90773-REDT, and RED2018-102641-T) and the Galician Ministry of Education, University and Professional Training (grants ED431F 2018/02, ED431C 2018/29, ED431G/08, and ED431G2019/04), with all grants co-funded by the European Regional Development Fund (ERDF/FEDER program) SI

The prevalence of Alzheimer’s disease (AD) in the growing elderly population makes accurately predicting AD progression crucial. Due to AD’s complex etiology and pathogenesis, an effective and medically practical solution is a challenging task. In this paper, we developed and evaluated two novel hybrid deep learning architectures for AD progression detection. These models are based on the fusion of multiple deep bidirectional long short-term memory (BiLSTM) models. The first architecture is an interpretable multitask regression model that predicts seven crucial cognitive scores for the patient 2.5 years after their last observations. The predicted scores are used to build an interpretable clinical decision support system based on a glass-box model. This architecture aims to explore the role of multitasking models in producing more stable, robust, and accurate results. The second architecture is a hybrid model where the deep features extracted from the BiLSTM model are used to train multiple machine learning classifiers. The two architectures were comprehensively evaluated using different time series modalities of 1371 subjects participated in the study of the Alzheimer’s disease neuroimaging initiative (ADNI). The extensive, real-world experimental results over ADNI data help establish the effectiveness and practicality of the proposed deep learning models Dr. Jose M. Alonso is Ramon y Cajal Researcher (RYC-2016-19802), and its research is supported by the Spanish Ministry of Science, Innovation, and Universities (grants RTI2018-099646-B-I00, TIN2017-84796-C2-1-R, TIN2017-90773-REDT, and RED2018-102641-T) and the Galician Ministry of Education, University and Professional Training (grants ED431F 2018/02, ED431C 2018/29, ED431G/08, and ED431G2019/04), with all grants co-funded by the European Regional Development Fund (ERDF/FEDER program). This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (NRF-2016R1D1A1A03934816) 2022-12-25 SI

In this chapter, we describe how to generate not only interpretable but also self-explaining fuzzy systems. Such systems are expected to manage information granules naturally as humans do. We take as starting point the Fuzzy Unordered Rule Induction Algorithm (FURIA for short) which produces a good interpretability-accuracy trade-off. FURIA rules have local semantics and manage information granules without linguistic interpretability. With the aim of making FURIA rules self-explaining, we have created a linguistic layer which endows FURIA with global semantics and linguistic interpretability. Explainable FURIA rules provide users with evidence-based (factual) and counterfactual explanations for single classifications. Factual explanations answer the question why a particular class is selected in terms of the given observations. In addition, counterfactual explanations pay attention to why the rest of classes are not selected. Thus, endowing FURIA rules with the capability to generate a combination of both factual and counterfactual explanations is likely to make them more trustworthy. We illustrate how to build self-explaining FURIA classifiers in two practical use cases regarding beer style classification and vehicle classification. Experimental results are encouraging. The generated classifiers exhibit accuracy comparable to a black-box classifier such as Random Forest. Moreover, their explainability is comparable to that provided by white-box classifiers designed with the Highly Interpretable Linguistic Knowledge fuzzy modeling methodology (HILK for short) in terms of explainability.