INSAFEDARE Project: Innovative Applications of Assessment and Assurance of Data and Synthetic Data for Regulatory Decision Support.

Digital health solutions hold promise for enhancing healthcare delivery and patient outcomes, primarily driven by advancements such as machine learning, artificial intelligence, and data science, which enable the development of integrated care systems. Techniques for generating synthetic data from real datasets are highly advanced and continually evolving. This paper aims to present the INSAFEDARE project's ambition regarding medical devices' regulation and how real and synthetic data can be used to check if devices are safe and effective. The project will consist of three pillars: a) assurance of new state-of-the-art technologies and approaches (such as synthetic data), which will support the validation methods as part of regulatory decision-making; b) technical and scientific, focusing on data-based safety assurance, as well as discovery, integration and use of datasets, and use of machine learning approaches; and c) delivery to practice, through co-production involving relevant stakeholders, dissemination and sustainability of the project's outputs. Finally, INSAFEDARE will develop an open syllabus and training certification for health professionals focused on quality assurance.

On the evaluation of synthetic longitudinal electronic health records.

BACKGROUND: Synthetic Electronic Health Records (EHRs) are becoming increasingly popular as a privacy enhancing technology. However, for longitudinal EHRs specifically, little research has been done into how to properly evaluate synthetically generated samples. In this article, we provide a discussion on existing methods and recommendations when evaluating the quality of synthetic longitudinal EHRs. METHODS: We recommend to assess synthetic EHR quality through similarity to real EHRs in low-dimensional projections, accuracy of a classifier discriminating synthetic from real samples, performance of synthetic versus real trained algorithms in clinical tasks, and privacy risk through risk of attribute inference. For each metric we discuss strengths and weaknesses, next to showing how it can be applied on a longitudinal dataset. RESULTS: To support the discussion on evaluation metrics, we apply discussed metrics on a dataset of synthetic EHRs generated from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) repository. CONCLUSIONS: The discussion on evaluation metrics provide guidance for researchers on how to use and interpret different metrics when evaluating the quality of synthetic longitudinal EHRs.

Extracting patient lifestyle characteristics from Dutch clinical text with BERT models.

BACKGROUND: BERT models have seen widespread use on unstructured text within the clinical domain. However, little to no research has been conducted into classifying unstructured clinical notes on the basis of patient lifestyle indicators, especially in Dutch. This article aims to test the feasibility of deep BERT models on the task of patient lifestyle classification, as well as introducing an experimental framework that is easily reproducible in future research. METHODS: This study makes use of unstructured general patient text data from HagaZiekenhuis, a large hospital in The Netherlands. Over 148 000 notes were provided to us, which were each automatically labelled on the basis of the respective patients' smoking, alcohol usage and drug usage statuses. In this paper we test feasibility of automatically assigning labels, and justify it using hand-labelled input. Ultimately, we compare macro F1-scores of string matching, SGD and several BERT models on the task of classifying smoking, alcohol and drug usage. We test Dutch BERT models and English models with translated input. RESULTS: We find that our further pre-trained MedRoBERTa.nl-HAGA model outperformed every other model on smoking (0.93) and drug usage (0.77). Interestingly, our ClinicalBERT model that was merely fine-tuned on translated text performed best on the alcohol task (0.80). In t-SNE visualisations, we show our MedRoBERTa.nl-HAGA model is the best model to differentiate between classes in the embedding space, explaining its superior classification performance. CONCLUSIONS: We suggest MedRoBERTa.nl-HAGA to be used as a baseline in future research on Dutch free text patient lifestyle classification. We furthermore strongly suggest further exploring the application of translation to input text in non-English clinical BERT research, as we only translated a subset of the full set and yet achieved very promising results.

Widespread chromatin context-dependencies of DNA double-strand break repair proteins.

DNA double-strand breaks are repaired by multiple pathways, including non-homologous end-joining (NHEJ) and microhomology-mediated end-joining (MMEJ). The balance of these pathways is dependent on the local chromatin context, but the underlying mechanisms are poorly understood. By combining knockout screening with a dual MMEJ:NHEJ reporter inserted in 19 different chromatin environments, we identified dozens of DNA repair proteins that modulate pathway balance dependent on the local chromatin state. Proteins that favor NHEJ mostly synergize with euchromatin, while proteins that favor MMEJ generally synergize with distinct types of heterochromatin. Examples of the former are BRCA2 and POLL, and of the latter the FANC complex and ATM. Moreover, in a diversity of human cancer types, loss of several of these proteins alters the distribution of pathway-specific mutations between heterochromatin and euchromatin. Together, these results uncover a complex network of proteins that regulate MMEJ:NHEJ balance in a chromatin context-dependent manner.

High-throughput identification of human SNPs affecting regulatory element activity.

Most of the millions of SNPs in the human genome are non-coding, and many overlap with putative regulatory elements. Genome-wide association studies (GWAS) have linked many of these SNPs to human traits or to gene expression levels, but rarely with sufficient resolution to identify the causal SNPs. Functional screens based on reporter assays have previously been of insufficient throughput to test the vast space of SNPs for possible effects on regulatory element activity. Here we leveraged the throughput and resolution of the survey of regulatory elements (SuRE) reporter technology to survey the effect of 5.9 million SNPs, including 57% of the known common SNPs, on enhancer and promoter activity. We identified more than 30,000 SNPs that alter the activity of putative regulatory elements, partially in a cell-type-specific manner. Integration of this dataset with GWAS results may help to pinpoint SNPs that underlie human traits.

Kinetics and Fidelity of the Repair of Cas9-Induced Double-Strand DNA Breaks.

The RNA-guided DNA endonuclease Cas9 is a powerful tool for genome editing. Little is known about the kinetics and fidelity of the double-strand break (DSB) repair process that follows a Cas9 cutting event in living cells. Here, we developed a strategy to measure the kinetics of DSB repair for single loci in human cells. Quantitative modeling of repaired DNA in time series after Cas9 activation reveals variable and often slow repair rates, with half-life times up to ∼10 hr. Furthermore, repair of the DSBs tends to be error prone. Both classical and microhomology-mediated end joining pathways contribute to the erroneous repair. Estimation of their individual rate constants indicates that the balance between these two pathways changes over time and can be altered by additional ionizing radiation. Our approach provides quantitative insights into DSB repair kinetics and fidelity in single loci and indicates that Cas9-induced DSBs are repaired in an unusual manner.

Massive reshaping of genome-nuclear lamina interactions during oncogene-induced senescence.

Cellular senescence is a mechanism that virtually irreversibly suppresses the proliferative capacity of cells in response to various stress signals. This includes the expression of activated oncogenes, which causes Oncogene-Induced Senescence (OIS). A body of evidence points to the involvement in OIS of chromatin reorganization, including the formation of senescence-associated heterochromatic foci (SAHF). The nuclear lamina (NL) is an important contributor to genome organization and has been implicated in cellular senescence and organismal aging. It interacts with multiple regions of the genome called lamina-associated domains (LADs). Some LADs are cell-type specific, whereas others are conserved between cell types and are referred to as constitutive LADs (cLADs). Here, we used DamID to investigate the changes in genome-NL interactions in a model of OIS triggered by the expression of the common BRAFV600E oncogene. We found that OIS cells lose most of their cLADS, suggesting the loss of a specific mechanism that targets cLADs to the NL. In addition, multiple genes relocated to the NL. Unexpectedly, they were not repressed, implying the abrogation of the repressive activity of the NL during OIS. Finally, OIS cells displayed an increased association of telomeres with the NL. Our study reveals that senescent cells acquire a new type of LAD organization and suggests the existence of as yet unknown mechanisms that tether cLADs to the NL and repress gene expression at the NL.

Genome-wide mapping of autonomous promoter activity in human cells.

Previous methods to systematically characterize sequence-intrinsic activity of promoters have been limited by relatively low throughput and the length of the sequences that could be tested. Here we present 'survey of regulatory elements' (SuRE), a method that assays more than 108 DNA fragments, each 0.2-2 kb in size, for their ability to drive transcription autonomously. In SuRE, a plasmid library of random genomic fragments upstream of a 20-bp barcode is constructed, and decoded by paired-end sequencing. This library is used to transfect cells, and barcodes in transcribed RNA are quantified by high-throughput sequencing. When applied to the human genome, we achieve 55-fold genome coverage, allowing us to map autonomous promoter activity genome-wide in K562 cells. By computational modeling we delineate subregions within promoters that are relevant for their activity. We show that antisense promoter transcription is generally dependent on the sense core promoter sequences, and that most enhancers and several families of repetitive elements act as autonomous transcription initiation sites.

ATP-binding Cassette Subfamily C Member 5 (ABCC5) Functions as an Efflux Transporter of Glutamate Conjugates and Analogs.

The ubiquitous efflux transporter ABCC5 (ATP-binding cassette subfamily C member 5) is present at high levels in the blood-brain barrier, neurons, and glia, but its in vivo substrates and function are not known. Using untargeted metabolomic screens, we show that Abcc5(-/-) mice accumulate endogenous glutamate conjugates in several tissues, but brain in particular. The abundant neurotransmitter N-acetylaspartylglutamate was 2.4-fold higher in Abcc5(-/-) brain. The metabolites that accumulated in Abcc5(-/-) tissues were depleted in cultured cells that overexpressed human ABCC5. In a vesicular membrane transport assay, ABCC5 also transported exogenous glutamate analogs, like the classic excitotoxic neurotoxins kainic acid, domoic acid, and NMDA; the therapeutic glutamate analog ZJ43; and, as previously shown, the anti-cancer drug methotrexate. Glutamate conjugates and analogs are of physiological relevance because they can affect the function of glutamate, the principal excitatory neurotransmitter in the brain. After CO2 asphyxiation, several immediate early genes were expressed at lower levels in Abcc5(-/-) brains than in wild type brains, suggesting altered glutamate signaling. Our results show that ABCC5 is a general glutamate conjugate and analog transporter that affects the disposition of endogenous metabolites, toxins, and drugs.

ABCC6 prevents ectopic mineralization seen in pseudoxanthoma elasticum by inducing cellular nucleotide release.

Pseudoxanthoma elasticum (PXE) is an autosomal recessive disease characterized by progressive ectopic mineralization of the skin, eyes, and arteries, for which no effective treatment exists. PXE is caused by inactivating mutations in the gene encoding ATP-binding cassette sub-family C member 6 (ABCC6), an ATP-dependent efflux transporter present mainly in the liver. Abcc6(-/-) mice have been instrumental in demonstrating that PXE is a metabolic disease caused by the absence of an unknown factor in the circulation, the presence of which depends on ABCC6 in the liver. Why absence of this factor results in PXE has remained a mystery. Here we report that medium from HEK293 cells overexpressing either human or rat ABCC6 potently inhibits mineralization in vitro, whereas medium from HEK293 control cells does not. Untargeted metabolomics revealed that cells expressing ABCC6 excrete large amounts of nucleoside triphosphates, even though ABCC6 itself does not transport nucleoside triphosphates. Extracellularly, ectonucleotidases hydrolyze the excreted nucleoside triphosphates to nucleoside monophosphates and inorganic pyrophosphate (PPi), a strong inhibitor of mineralization that plays a pivotal role in several mineralization disorders similar to PXE. The in vivo relevance of our data are demonstrated in Abcc6(-/-) mice, which had plasma PPi levels <40% of those found in WT mice. This study provides insight into how ABCC6 affects PXE. Our data indicate that the factor that normally prevents PXE is PPi, which is provided to the circulation in the form of nucleoside triphosphates via an as-yet unidentified but ABCC6-dependent mechanism.

A network model of the molecular organization of chromatin in Drosophila.

Chromatin governs gene regulation and genome maintenance, yet a substantial fraction of the chromatin proteome is still unexplored. Moreover, a global model of the chromatin protein network is lacking. By screening >100 candidates we identify 42 Drosophila proteins that were not previously associated with chromatin, which all display specific genomic binding patterns. Bayesian network modeling of the binding profiles of these and 70 known chromatin components yields a detailed blueprint of the in vivo chromatin protein network. We demonstrate functional compartmentalization of this network, and predict functions for most of the previously unknown chromatin proteins, including roles in DNA replication and repair, and gene activation and repression.

Transportomics: screening for substrates of ABC transporters in body fluids using vesicular transport assays.

The ATP-binding cassette (ABC) genes encode the largest family of transmembrane proteins. ABC transporters translocate a wide variety of substrates across membranes, but their physiological function is often incompletely understood. We describe a new method to study the substrate spectrum of ABC transporters: We incubate extracts of mouse urine with membrane vesicles prepared from Spodoptera frugiperda Sf9 insect cells overproducing an ABC transporter and determine the compounds transported into the vesicles by LC/MS-based metabolomics. We illustrate the power of this simple "transportomics" approach using ABCC2, a protein present at sites of uptake and elimination. We identified many new substrates of ABCC2 in urine. These included glucuronides of plant-derived xenobiotics, a class of compounds to which humans are exposed on a daily basis. Moreover, we show that the excretion of these compounds in vivo depends on ABCC2: compared to wild-type mice, the urinary excretion of several glucuronides was increased up to 20-fold in Abcc2(-/-) mice. Transportomics has broad applicability, as it is not restricted to urine and can be applied to other ATP-dependent transport proteins as well.

Contribution of the drug transporter ABCG2 (breast cancer resistance protein) to resistance against anticancer nucleosides.

We have studied the potential contribution of ABCG2 (breast cancer resistance protein) to resistance to nucleoside analogues. In cells transfected with DNA constructs resulting in overexpression of human or mouse ABCG2, we found resistance against cladribine, clofarabine, fludarabine, 6-mercaptopurine, and 6-mercaptopurine riboside in both MDCKII and HEK293 cells and against gemcitabine only in HEK293 cells. With Transwell studies in MDCK cells and transport experiments with vesicles from Sf9 and HEK293 cells, we show that ABCG2 is able to transport not only the nucleotide CdAMP, like several other ATP-binding cassette transporters of the ABCC (multidrug resistance protein) family, but also the nucleoside cladribine itself. Expression of ABCG2 in cells results in a substantial decrease of intracellular CdATP, explaining the resistance against cladribine. The high transport rate of cladribine and clofarabine by ABCG2 deduced from Transwell experiments raises the possibility that this transporter could affect the disposition of nucleoside analogues in patients or cause resistance in tumors.

Multidrug resistance-associated protein 9 (ABCC12) is present in mouse and boar sperm.

The human and murine genes for MRP9 (multidrug resistance-associated protein 9; ABCC12) yield many alternatively spliced RNAs. Using a panel of monoclonal antibodies, we detected full-length Mrp9 only in testicular germ cells and mouse sperm; we obtained no evidence for the existence of the truncated 100 kDa MRP9 protein reported previously. In contrast with other MRPs, neither murine Mrp9 nor the human MRP9 produced in MRP9-transfected HEK-293 cells (human embryonic kidney cells) appears to contain N-linked carbohydrates. In mouse and boar sperm, Mrp9 localizes to the midpiece, a structure containing all sperm mitochondria. However, immunolocalization microscopy and cell fractionation studies with transfected HEK-293 cells and mouse testis show that MRP9/Mrp9 does not localize to mitochondria. In HEK-293 cells, it is predominantly localized in the endoplasmic reticulum. We have been unable to demonstrate transport by MRP9 of substrates transported by other MRPs, such as drug conjugates and other organic anions.

cGMP transport by vesicles from human and mouse erythrocytes.

cGMP secretion from cells can be mediated by ATP-binding cassette (ABC) transporters ABCC4, ABCC5, and ABCC11. Indirect evidence suggests that ABCC4 and ABCC5 contribute to cGMP transport by erythrocytes. We have re-investigated the issue using erythrocytes from wild-type and transporter knockout mice. Murine wild-type erythrocyte vesicles transported cGMP with an apparent Km that was 100-fold higher than their human counterparts, the apparent Vmax being similar. Whereas cGMP transport into human vesicles was efficiently inhibited by the ABCC4-specific substrate prostaglandin E1, cGMP transport into mouse vesicles was inhibited equally by Abcg2 and Abcc4 inhibitors/substrates. Similarly, cGMP transport into vesicles from Abcc4-/- and Abcg2-/- mice was 42% and 51% of that into wild-type mouse vesicles, respectively, whereas cGMP transport into vesicles from Abcc4(-/-)/Abcg2(-/-) mice was near background. The knockout mice were used to show that Abcg2-mediated cGMP transport occurred with lower affinity but higher Vmax than Abcc4-mediated transport. Involvement of Abcg2 in cGMP transport by Abcc4-/- erythrocyte vesicles was supported by higher transport at pH 5.5 than at pH 7.4, a characteristic of Abcg2-mediated transport. The relative contribution of ABCC4/Abcc4 and ABCG2/Abcg2 in cGMP transport was confirmed with a new inhibitor of ABCC4 transport, the protease inhibitor 4-(2-aminoethyl)benzenesulfonyl fluoride.

The human multidrug resistance protein MRP5 transports folates and can mediate cellular resistance against antifolates.

Members of the multidrug resistance protein family, notably MRP1-4/ABCC1-4, and the breast cancer resistance protein BCRP/ABCG2 have been recognized as cellular exporters for the folate antagonist methotrexate (MTX). Here we show that MRP5/ABCC5 is also an antifolate and folate exporter based on the following evidence: (a) Using membrane vesicles from HEK293 cells, we show that MRP5 transports both MTX (KM = 1.3 mmol/L and VMAX = 780 pmol per mg protein per minute) and folic acid (KM = 1.0 mmol/L and VMAX = 875 pmol per mg protein per minute). MRP5 also transports MTX-glu2 (KM = 0.7 mmol/L and VMAX = 450 pmol per mg protein per minute) but not MTX-glu3. (b) Both accumulation of total [3H]MTX and of MTX polyglutamates were significantly reduced in MRP5 overexpressing cells. (c) Cell growth inhibition studies with MRP5 transfected HEK293 cells showed that MRP5 conferred high-level resistance (>160-fold) against the antifolates MTX, GW1843, and ZD1694 (raltitrexed) in short-term (4 hours) incubations with high drug concentrations; this resistance was proportional to the MRP5 level. (d) MRP5-mediated resistance (8.5- and 2.1-fold) was also found in standard long-term incubations (72 hours) at low concentrations of ZD1694 and GW1843. These results show the potential of MRP5 to mediate transport of (anti)folates and contribute to resistance against antifolate drugs.

The human multidrug resistance protein MRP4 functions as a prostaglandin efflux transporter and is inhibited by nonsteroidal antiinflammatory drugs.

Prostaglandins are involved in a wide variety of physiological and pathophysiological processes, but the mechanism of prostaglandin release from cells is not completely understood. Although poorly membrane permeable, prostaglandins are believed to exit cells by passive diffusion. We have investigated the interaction between prostaglandins and members of the ATP-binding cassette (ABC) transporter ABCC [multidrug resistance protein (MRP)] family of membrane export pumps. In inside-out membrane vesicles derived from insect cells or HEK293 cells, MRP4 catalyzed the time- and ATP-dependent uptake of prostaglandin E1 (PGE1) and PGE2. In contrast, MRP1, MRP2, MRP3, and MRP5 did not transport PGE1 or PGE2. The MRP4-mediated transport of PGE1 and PGE2 displayed saturation kinetics, with Km values of 2.1 and 3.4 microM, respectively. Further studies showed that PGF1alpha, PGF2alpha, PGA1, and thromboxane B2 were high-affinity inhibitors (and therefore presumably substrates) of MRP4. Furthermore, several nonsteroidal antiinflammatory drugs were potent inhibitors of MRP4 at concentrations that did not inhibit MRP1. In cells expressing the prostaglandin transporter PGT, the steady-state accumulation of PGE1 and PGE2 was reduced proportional to MRP4 expression. Inhibition of MRP4 by an MRP4-specific RNA interference construct or by indomethacin reversed this accumulation deficit. Together, these data suggest that MRP4 can release prostaglandins from cells, and that, in addition to inhibiting prostaglandin synthesis, some nonsteroidal antiinflammatory drugs might also act by inhibiting this release.

Characterization of the transport of nucleoside analog drugs by the human multidrug resistance proteins MRP4 and MRP5.

The human multidrug resistance proteins MRP4 and MRP5 are organic anion transporters that have the unusual ability to transport cyclic nucleotides and some nucleoside monophosphate analogs. Base and nucleoside analogs used in the chemotherapy of cancer and viral infections are potential substrates. To assess the possible contribution of MRP4 and MRP5 to resistance against these drugs, we have investigated the transport mediated by MRP4 and MRP5. In cytotoxicity assays, MRP4 conferred resistance to the antiviral agent 9-(2-phosphonomethoxyethyl)adenine (PMEA) and high-performance liquid chromatography analysis showed that, like MRP5, MRP4 transported PMEA in an unmodified form. MRP4 also mediated substantial resistance against other acyclic nucleoside phosphonates, whereas MRP5 did not. Apart from low-level MRP4-mediated cladribine resistance, the cytotoxicity of clinically used anticancer nucleosides was not influenced by overexpression of MRP4 or MRP5. In contrast, MRP5 mediated efflux of the pyrimidine-based antiviral 2',3'-dideoxynucleoside 2',3'-didehydro-2',3'-dideoxythymidine 5'-monophosphate (d4TMP) and its phosphoramidate derivative alaninyl-d4TMP from cells loaded with the 2',3'-didehydro-2',3'-dideoxythymidine prodrugs cyclosaligenyl-d4TMP and aryloxyphosphoramidate d4TMP (So324), respectively. Moreover, only inside-out membrane vesicles derived from MRP5-overexpressing cells accumulated alaninyl-d4TMP. Cellular efflux and vesicular uptake studies were carried out to further compare transport mediated by MRP4 and MRP5 and showed that dipyridamole, dilazep, nitrobenzyl mercaptopurine riboside, sildenafil, trequinsin and MK571 inhibited MRP4 more than MRP5, whereas cyclic nucleotides and monophosphorylated nucleoside analogs were equally poor inhibitors of both pumps. These results strongly suggest that the affinity of MRP4 and MRP5 for nucleotide-based substrates is low.