Understanding metric-related pitfalls in image analysis validation.

Validation metrics are key for tracking scientific progress and bridging the current chasm between artificial intelligence research and its translation into practice. However, increasing evidence shows that, particularly in image analysis, metrics are often chosen inadequately. Although taking into account the individual strengths, weaknesses and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multistage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides a reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Although focused on biomedical image analysis, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. The work serves to enhance global comprehension of a key topic in image analysis validation.

Metrics reloaded: recommendations for image analysis validation.

Increasing evidence shows that flaws in machine learning (ML) algorithm validation are an underestimated global problem. In biomedical image analysis, chosen performance metrics often do not reflect the domain interest, and thus fail to adequately measure scientific progress and hinder translation of ML techniques into practice. To overcome this, we created Metrics Reloaded, a comprehensive framework guiding researchers in the problem-aware selection of metrics. Developed by a large international consortium in a multistage Delphi process, it is based on the novel concept of a problem fingerprint-a structured representation of the given problem that captures all aspects that are relevant for metric selection, from the domain interest to the properties of the target structure(s), dataset and algorithm output. On the basis of the problem fingerprint, users are guided through the process of choosing and applying appropriate validation metrics while being made aware of potential pitfalls. Metrics Reloaded targets image analysis problems that can be interpreted as classification tasks at image, object or pixel level, namely image-level classification, object detection, semantic segmentation and instance segmentation tasks. To improve the user experience, we implemented the framework in the Metrics Reloaded online tool. Following the convergence of ML methodology across application domains, Metrics Reloaded fosters the convergence of validation methodology. Its applicability is demonstrated for various biomedical use cases.

DNA choreography: correlating mobility and organization of DNA across different resolutions from loops to chromosomes.

The dynamics of DNA in the cell nucleus plays a role in cellular processes and fates but the interplay of DNA mobility with the hierarchical levels of DNA organization is still underexplored. Here, we made use of DNA replication to directly label genomic DNA in an unbiased genome-wide manner. This was followed by live-cell time-lapse microscopy of the labeled DNA combining imaging at different resolutions levels simultaneously and allowing one to trace DNA motion across organization levels within the same cells. Quantification of the labeled DNA segments at different microscopic resolution levels revealed sizes comparable to the ones reported for DNA loops using 3D super-resolution microscopy, topologically associated domains (TAD) using 3D widefield microscopy, and also entire chromosomes. By employing advanced chromatin tracking and image registration, we discovered that DNA exhibited higher mobility at the individual loop level compared to the TAD level and even less at the chromosome level. Additionally, our findings indicate that chromatin movement, regardless of the resolution, slowed down during the S phase of the cell cycle compared to the G1/G2 phases. Furthermore, we found that a fraction of DNA loops and TADs exhibited directed movement with the majority depicting constrained movement. Our data also indicated spatial mobility differences with DNA loops and TADs at the nuclear periphery and the nuclear interior exhibiting lower velocity and radius of gyration than the intermediate locations. On the basis of these insights, we propose that there is a link between DNA mobility and its organizational structure including spatial distribution, which impacts cellular processes.

Probing protein ubiquitination in live cells.

The reversible attachment of ubiquitin governs the interaction, activity and degradation of proteins whereby the type and target of this conjugation determine the biological response. The investigation of this complex and multi-faceted protein ubiquitination mostly relies on painstaking biochemical analyses. Here, we employ recombinant binding domains to probe the ubiquitination of proteins in living cells. We immobilize GFP-fused proteins of interest at a distinct cellular structure and detect their ubiquitination state with red fluorescent ubiquitin binders. With this ubiquitin fluorescent three-hybrid (ubiF3H) assay we identified HP1ß as a novel ubiquitination target of UHRF1. The use of linkage specific ubiquitin binding domains enabled the discrimination of K48 and K63 linked protein ubiquitination. To enhance signal-to-noise ratio, we implemented fluorescence complementation (ubiF3Hc) with split YFP. Using in addition a cell cycle marker we could show that HP1ß is mostly ubiquitinated by UHRF1 during S phase and deubiquitinated by the protease USP7. With this complementation assay we could also directly detect the ubiquitination of the tumor suppressor p53 and monitor its inhibition by the anti-cancer drug Nutlin-3. Altogether, we demonstrate the utility of the ubiF3H assay to probe the ubiquitination of specific proteins and to screen for ligases, proteases and small molecules controlling this posttranslational modification.

Omicron - The new SARS-CoV-2 challenge?

SARS-CoV-2 virus has infected nearly 300 M people worldwide and has been associated with over 6 M deaths by March 2022. Since the virus emergence in December 2019 in Wuhan, several new mutations have been described. The World Health Organization has developed a working name for these emerging variants according to their impact on the worldwide population. In this context a high alert has been paid to variants of concern (VOC) due to their high infectiousness and transmissibility patterns. The most recent VOC, Omicron (B.1.1.529), has become dominant in the shortest time ever and has placed Europe under an overwhelming and unprecedented number of new cases. This variant has numerous mutations in regions that are associated with higher transmissibility, stronger viral binding, affinity and antibody escape. Moreover, the mutations and deletions present in the spike protein suggest that the SARS-CoV-2 specific attachment inhibitors may not be the best option for Omicron therapy. Omicron is the dominant variant circulating worldwide and, at the end of February 2022, it was responsible for nearly all sequences reported to GISAID. Omicron is made up of several sublineages, where the most common are BA.1 and BA.2 (or Nextstrain clade 21K and 21L, respectively). At a global level, it is possible to say that the proportion of BA.2 has been increasing relative to BA.1 and in some countries it has been replacing it at high rates. In order to better assess the Omicron effectiveness on antibody escape, spread and infectious ability it is of the highest relevance to maintain a worldwide tight surveillance. Even though this variant has been associated with a lower death rate, it is important to highlight that the number of people becoming infected is concerning and that further unpredictable mutations may emerge as the number of infected people rises.

Visualization and characterization of RNA-protein interactions in living cells.

RNA-protein interactions are the structural and functional basis of significant numbers of RNA molecules. RNA-protein interaction assays though, still mainly depend on biochemical tests in vitro. Here, we establish a convenient and reliable RNA fluorescent three-hybrid (rF3H) method to detect/interrogate the interactions between RNAs and proteins in cells. A GFP tagged highly specific RNA trap is constructed to anchor the RNA of interest to an artificial or natural subcellular structure, and RNA-protein interactions can be detected and visualized by the enrichment of RNA binding proteins (RBPs) at these structures. Different RNA trapping systems are developed and detection of RNA-protein complexes at multiple subcellular structures are assayed. With this new toolset, interactions between proteins and mRNA or noncoding RNAs are characterized, including the interaction between a long noncoding RNA and an epigenetic modulator. Our approach provides a flexible and reliable method for the characterization of RNA-protein interactions in living cells.

Cytosine base modifications regulate DNA duplex stability and metabolism.

DNA base modifications diversify the genome and are essential players in development. Yet, their influence on DNA physical properties and the ensuing effects on genome metabolism are poorly understood. Here, we focus on the interplay of cytosine modifications and DNA processes. We show by a combination of in vitro reactions with well-defined protein compositions and conditions, and in vivo experiments within the complex networks of the cell that cytosine methylation stabilizes the DNA helix, increasing its melting temperature and reducing DNA helicase and RNA/DNA polymerase speed. Oxidation of methylated cytosine, however, reverts the duplex stabilizing and genome metabolic effects to the level of unmodified cytosine. We detect this effect with DNA replication and transcription proteins originating from different species, ranging from prokaryotic and viral to the eukaryotic yeast and mammalian proteins. Accordingly, lack of cytosine methylation increases replication fork speed by enhancing DNA helicase unwinding speed in cells. We further validate that this cannot simply be explained by altered global DNA decondensation, changes in histone marks or chromatin structure and accessibility. We propose that the variegated deposition of cytosine modifications along the genome regulates DNA helix stability, thereby providing an elementary mechanism for local fine-tuning of DNA metabolism.

Phosphorylation of the HP1ß hinge region sequesters KAP1 in heterochromatin and promotes the exit from naïve pluripotency.

Heterochromatin binding protein HP1ß plays an important role in chromatin organization and cell differentiation, however the underlying mechanisms remain unclear. Here, we generated HP1ß-/- embryonic stem cells and observed reduced heterochromatin clustering and impaired differentiation. We found that during stem cell differentiation, HP1ß is phosphorylated at serine 89 by CK2, which creates a binding site for the pluripotency regulator KAP1. This phosphorylation dependent sequestration of KAP1 in heterochromatin compartments causes a downregulation of pluripotency factors and triggers pluripotency exit. Accordingly, HP1ß-/- and phospho-mutant cells exhibited impaired differentiation, while ubiquitination-deficient KAP1-/- cells had the opposite phenotype with enhanced differentiation. These results suggest that KAP1 regulates pluripotency via its ubiquitination activity. We propose that the formation of subnuclear membraneless heterochromatin compartments may serve as a dynamic reservoir to trap or release cellular factors. The sequestration of essential regulators defines a novel and active role of heterochromatin in gene regulation and represents a dynamic mode of remote control to regulate cellular processes like cell fate decisions.

COVID-19 vaccines effectiveness against symptomatic disease and severe outcomes, 2021-2022: a test-negative case-control study.

OBJECTIVES: This study evaluated the effectiveness of COVID-19 vaccines in preventing symptomatic and severe disease. STUDY DESIGN: This was an observational test-negative case-control study. METHODS: Study participants were adults with at least one symptom included in the World Health Organization COVID-19 definition who sought health care in a public emergency department between 1 November 2021 and 2 March 2022 (corresponding with the fifth pandemic wave in Portugal dominated by the Omicron variant). This study used multivariable logistic regression models to estimate and compare the odds ratio of vaccination between test-positive cases and test-negative controls to calculate the absolute and relative vaccine effectiveness. RESULTS: The study included 1059 individuals (522 cases and 537 controls) with a median age of 56 years and 58% were women. Compared with the effectiveness of the primary vaccination scheme that had been completed ≥180 days earlier, the relative effectiveness against symptomatic infection of a booster administered between 14 and 132 days earlier was 71% (95% confidence interval [CI]: 57%, 81%; P < 0.001). The effectiveness of the primary series against symptomatic infection peaked at 85% (95% CI: 56%, 95%) between 14 and 90 days after the last inoculation and decreased to 34% (95% CI: -43%, 50%) after ≥180 days. CONCLUSIONS: Despite the known immunological evasion characteristics of the Omicron variant, results from this study show that vaccine effectiveness increases after booster administration. COVID-19 vaccine effectiveness decreases to less than 50% between 3 and 6 months after completion of the primary cycle; therefore, this would be an appropriate time to administer a booster to restore immunity.

ESO-ESMO fifth international consensus guidelines for breast cancer in young women (BCY5).

We dedicate this manuscript in memory of a dear friend and colleague Bella Kaufman. The fifth International Consensus Symposium for Breast Cancer in Young Women (BCY5) took place virtually in October 2020, organized by the European School of Oncology (ESO) and the European Society of Medical Oncology (ESMO). Consensus recommendations for the management of breast cancer in young women were updated from BCY4 with incorporation of new evidence to inform the guidelines. Areas of research priorities as well as specificities in different geographic and minority populations were identified. This manuscript summarizes the ESO-ESMO international consensus recommendations, which are also endorsed by the European Society of Breast Specialists (EUSOMA).

Validation strategies for antibodies targeting modified ribonucleotides.

Chemical modifications are found on almost all RNAs and affect their coding and noncoding functions. The identification of m6A on mRNA and its important role in gene regulation stimulated the field to investigate whether additional modifications are present on mRNAs. Indeed, modifications including m1A, m5C, m7G, 2'-OMe, and Ψ were detected. However, since their abundances are low and tools used for their corroboration are often not well characterized, their physiological relevance remains largely elusive. Antibodies targeting modified nucleotides are often used but have limitations such as low affinity or specificity. Moreover, they are not always well characterized and due to the low abundance of the modification, particularly on mRNAs, generated data sets might resemble noise rather than specific modification patterns. Therefore, it is critical that the affinity and specificity is rigorously tested using complementary approaches. Here, we provide an experimental toolbox that allows for testing antibody performance prior to their use.

Application of the workload indicators of staffing need method to calculate the size of the medical staff at a maternity hospital in the state of Bahia, Brazil.

BACKGROUND: Functioning health systems require a health workforce (HWF) that is qualified, available, equitably distributed, and accessible to the entire population as the basis for guaranteeing access to health. There is a global HWF crisis, manifested in Brazil by unequal distribution of healthcare personnel, particularly in rural areas, urban peripheries, and other hard-to-reach communities, posing a major obstacle to guaranteeing access to health systems and services. Based on the above, calculating the size and analyzing the workloads of the medical staff in the Obstetrics Department (OD) and Urgent Care Center (UCC) in a state maternity hospital is relevant for designing improvements in the work processes and future strategies for recruiting, selecting, and retaining these workers at the hospital, in turn favoring improvement in the quality of care for women and children at the state level. This scenario motivated the study's design, in which the overall objective was to analyze the workload of staff physicians working in the Obstetrics Department and Urgent Care Center of a public maternity hospital in the state of Bahia, based on the WISN method. METHODS: This was an exploratory-descriptive intervention study with a quantitative approach and qualitative elements, using the methodological stages recommended by the WISN to calculate and analyze the workload of obstetricians working in the OD and UCC in the maternity hospital. RESULTS: The study found a deficit of 14 shift obstetricians at the hospital with a workload of 0.81. The study also found that the insufficient number of obstetricians at the hospital resulted from precarious hiring formats, idle medical positions, and poorly structured work processes and a shortage in the multidisciplinary staff. CONCLUSION: The research sought to contribute to the reduction of the gap in models and methodologies for the staffing of gynecologists and obstetricians in Bahia Maternity Hospitals, without covering the whole subject, but to demonstrate that the findings of the workload analysis and its validation could be useful in promoting and directing the design and implementation of interventions to improve the quality of the workload.

Developmental differences in genome replication program and origin activation.

To ensure error-free duplication of all (epi)genetic information once per cell cycle, DNA replication follows a cell type and developmental stage specific spatio-temporal program. Here, we analyze the spatio-temporal DNA replication progression in (un)differentiated mouse embryonic stem (mES) cells. Whereas telomeres replicate throughout S-phase, we observe mid S-phase replication of (peri)centromeric heterochromatin in mES cells, which switches to late S-phase replication upon differentiation. This replication timing reversal correlates with and depends on an increase in condensation and a decrease in acetylation of chromatin. We further find synchronous duplication of the Y chromosome, marking the end of S-phase, irrespectively of the pluripotency state. Using a combination of single-molecule and super-resolution microscopy, we measure molecular properties of the mES cell replicon, the number of replication foci active in parallel and their spatial clustering. We conclude that each replication nanofocus in mES cells corresponds to an individual replicon, with up to one quarter representing unidirectional forks. Furthermore, with molecular combing and genome-wide origin mapping analyses, we find that mES cells activate twice as many origins spaced at half the distance than somatic cells. Altogether, our results highlight fundamental developmental differences on progression of genome replication and origin activation in pluripotent cells.

Bactericidal and antioxidant effects of essential oils from Satureja montana L., Myristica fragrans H. and Cymbopogon flexuosus.

The extraction and characterization of the essential oils (EO) from Satureja montana L., Myristica fragrans H. and Cymbopogon flexuosus and the determination of their antibacterial and antioxidant activities were achieved. The EO were identified by gas chromatography/mass spectrometry and quantified by gas chromatography using a flame ionization detector. The antibacterial potential against Escherichia coli and Staphylococcus aureus was evaluated by cell susceptibility assays and by scanning electron microscopy. The antioxidant activity was evaluated by the 2,2-diphenyl-1-picrylhydrazyl assay, by ß-carotene bleaching and by determining the reducing power. Borneol (36·18%), γ-terpineol (12·66%) and carvacrol (11·07%) were the principal components in the EO from S. montana, and sabinene (49·23%) and α-pinene (13·81%) were found in the EO from M. fragrans. Geranial (59·66%) and neral (38·98%) isomers were the only major components in the EO from C. flexuosus. The EO from S. montana was effective against E. coli, with minimum inhibitory and bactericidal concentrations (MIC and MBC) of 6·25 µl ml-1 , whereas bactericidal potential against both was observed for the EO from M. fragrans; MIC = 6·25 µl ml-1 for S. aureus and MBC = 12·5 µl ml-1 for E. coli. A significant protective role on lipid substrates in the ß-carotene bleaching assay was seen for the EO from S. montana and M. fragrans. Overall, such EO can be promising agents against pathogenic bacteria and for protecting biomolecules during oxidative stress.

Antifungal and antiocratoxigenic potential of Alpinia speciosa and Cymbopogon flexuosus essential oils encapsulated in poly(lactic acid) nanofibres against Aspergillus fungi.

Essential oils encapsulated in a polymeric matrix can be used as an alternative method to control fungi and mycotoxins. The essential oils were extracted by hydrodistillation and characterized by gas chromatography. The nanofibres were produced from poly (acid lactic) (PLA) containing essential oils by the Solution Blow Spinning method. The antifungal and antimicotoxygenic properties were evaluated against Aspergillus ochraceus and Aspergillus westerdijkiae by the fumigation method. Terpinen-4-ol (20·23%), sabinene (20·18%), 1·8-cineole (16·69%) and γ-terpinene (11·03%) were the principal compounds present in the essential oil from Alpinia speciosa, whereas citral (97·67%) was dominant from Cymbopogon flexuosus. Microscopy images showed that the addition of essential oils caused an increase in the diameter of the nanofibres. The infrared spectroscopy results indicated the presence of essential oils in the PLA nanofibres. Differential scanning calorimetry curves also indicated the existence of interactions between the essential oils and polymeric macromolecules through their plasticizing action. The hydrophobic character of nanofibres was revealed by the contact angle technique. An antifungal effect was observed, the mycelial growths (3·25-100%) and the synthesis of ochratoxin A (25·94-100%) were inhibited by the presence of the nanofibres. The results suggest that bioactive nanofibres hold promise for application to control toxigenic fungi.

Antifungal activity of poly(ε-caprolactone) nanoparticles incorporated with Eucalyptus essential oils against Hemileia vastatrix.

Coffee (Coffea L.) is one of the main crops produced globally. Its contamination by the fungus Hemileia vastatrix Berkeley and Broome has been economically detrimental for producers. The objective of this work was to extract and characterize the essential oils from Eucalyptus citriodora Hook, Eucalyptus camaldulensis Dehn and Eucalyptus grandis Hill ex Maiden, produce and characterize nanoparticles containing these essential oils and evaluate the in vivo and in vitro antifungal activity of free and nanoencapsulated essential oils. The principal constituent of the essential oil from E. citriodora was citronellal; that from E. grandis was α-pinene; and that from E. camaldulensis was 1,8-cineol. The in vitro antifungal activity against the fungus H. vastatrix was 100% at a concentration of 1000 µl l-1 for all the oils and nanoparticles containing these natural products. The sizes of the nanoparticles produced with the essential oils from E. citriodora, E. camaldulensis and E. grandis were 402·13 nm, 275·33 nm and 328·5 nm, respectively, with surface charges of -11·8 mV, -9·24 mV and - 6·76 mV, respectively. Fourier transform infrared analyses proved that the encapsulation of essential oils occurred in the polymeric matrix of poly(ε-caprolactone). The incorporation of essential oils into biodegradable poly(ε-caprolactone) nanoparticles increased their efficiency as biofungicides in the fight against coffee rust, decreasing the severity of the disease by up to 90·75% after treatment with the nanoparticles containing the essential oil from E. grandis.

Mechanism for autoinhibition and activation of the MORC3 ATPase.

Microrchidia 3 (MORC3) is a human protein linked to autoimmune disorders, Down syndrome, and cancer. It is a member of a newly identified family of human ATPases with an uncharacterized mechanism of action. Here, we elucidate the molecular basis for inhibition and activation of MORC3. The crystal structure of the MORC3 region encompassing the ATPase and CW domains in complex with a nonhydrolyzable ATP analog demonstrates that the two domains are directly coupled. The extensive ATPase:CW interface stabilizes the protein fold but inhibits the catalytic activity of MORC3. Enzymatic, NMR, mutational, and biochemical analyses show that in the autoinhibited, off state, the CW domain sterically impedes binding of the ATPase domain to DNA, which in turn is required for the catalytic activity. MORC3 autoinhibition is released by disrupting the intramolecular ATPase:CW coupling through the competitive interaction of CW with histone H3 tail or by mutating the interfacial residues. Binding of CW to H3 leads to a marked rearrangement in the ATPase-CW cassette, which frees the DNA-binding site in active MORC3 (on state). We show that ATP-induced dimerization of the ATPase domain is strictly required for the catalytic activity and that the dimeric form of ATPase-CW might cooperatively bind to dsDNA. Together, our findings uncovered a mechanism underlying the fine-tuned regulation of the catalytic domain of MORC3 by the epigenetic reader, CW.

Applying causal models to explore the mechanism of action of simvastatin in progressive multiple sclerosis.

Understanding the mode of action of drugs is a challenge with conventional methods in clinical trials. Here, we aimed to explore whether simvastatin effects on brain atrophy and disability in secondary progressive multiple sclerosis (SPMS) are mediated by reducing cholesterol or are independent of cholesterol. We applied structural equation models to the MS-STAT trial in which 140 patients with SPMS were randomized to receive placebo or simvastatin. At baseline, after 1 and 2 years, patients underwent brain magnetic resonance imaging; their cognitive and physical disability were assessed on the block design test and Expanded Disability Status Scale (EDSS), and serum total cholesterol levels were measured. We calculated the percentage brain volume change (brain atrophy). We compared two models to select the most likely one: a cholesterol-dependent model with a cholesterol-independent model. The cholesterol-independent model was the most likely option. When we deconstructed the total treatment effect into indirect effects, which were mediated by brain atrophy, and direct effects, simvastatin had a direct effect (independent of serum cholesterol) on both the EDSS, which explained 69% of the overall treatment effect on EDSS, and brain atrophy, which, in turn, was responsible for 31% of the total treatment effect on EDSS [ß = -0.037; 95% credible interval (CI) = -0.075, -0.010]. This suggests that simvastatin's beneficial effects in MS are independent of its effect on lowering peripheral cholesterol levels, implicating a role for upstream intermediate metabolites of the cholesterol synthesis pathway. Importantly, it demonstrates that computational models can elucidate the causal architecture underlying treatment effects in clinical trials of progressive MS.

The Chromatin Architectural Protein CTCF Is Critical for Cell Survival upon Irradiation-Induced DNA Damage.

CTCF is a nuclear protein initially discovered for its role in enhancer-promoter insulation. It has been shown to play a role in genome architecture and in fact, its DNA binding sites are enriched at the borders of chromatin domains. Recently, we showed that depletion of CTCF impairs the DNA damage response to ionizing radiation. To investigate the relationship between chromatin domains and DNA damage repair, we present here clonogenic survival assays in different cell lines upon CTCF knockdown and ionizing irradiation. The application of a wide range of ionizing irradiation doses (0-10 Gy) allowed us to investigate the survival response through a biophysical model that accounts for the double-strand breaks' probability distribution onto chromatin domains. We demonstrate that the radiosensitivity of different cell lines is increased upon lowering the amount of the architectural protein. Our model shows that the deficiency in the DNA repair ability is related to the changes in the size of chromatin domains that occur when different amounts of CTCF are present in the nucleus.

Robust parametric modeling of Alzheimer's disease progression.

Quantitative characterization of disease progression using longitudinal data can provide long-term predictions for the pathological stages of individuals. This work studies the robust modeling of Alzheimer's disease progression using parametric methods. The proposed method linearly maps the individual's age to a disease progression score (DPS) and jointly fits constrained generalized logistic functions to the longitudinal dynamics of biomarkers as functions of the DPS using M-estimation. Robustness of the estimates is quantified using bootstrapping via Monte Carlo resampling, and the estimated inflection points of the fitted functions are used to temporally order the modeled biomarkers in the disease course. Kernel density estimation is applied to the obtained DPSs for clinical status classification using a Bayesian classifier. Different M-estimators and logistic functions, including a novel type proposed in this study, called modified Stannard, are evaluated on the data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) for robust modeling of volumetric magnetic resonance imaging (MRI) and positron emission tomography (PET) biomarkers, cerebrospinal fluid (CSF) measurements, as well as cognitive tests. The results show that the modified Stannard function fitted using the logistic loss achieves the best modeling performance with an average normalized mean absolute error (NMAE) of 0.991 across all biomarkers and bootstraps. Applied to the ADNI test set, this model achieves a multiclass area under the ROC curve (AUC) of 0.934 in clinical status classification. The obtained results for the proposed model outperform almost all state-of-the-art results in predicting biomarker values and classifying clinical status. Finally, the experiments show that the proposed model, trained using abundant ADNI data, generalizes well to data from the National Alzheimer's Coordinating Center (NACC) with an average NMAE of 1.182 and a multiclass AUC of 0.929.