Synaptic architecture of a memory engram in the mouse hippocampus.

Memory engrams are formed through experience-dependent remodeling of neural circuits, but their detailed architectures have remained unresolved. Using 3D electron microscopy, we performed nanoscale reconstructions of the hippocampal CA3-CA1 pathway following chemogenetic labeling of cellular ensembles with a remote history of correlated excitation during associative learning. Projection neurons involved in memory acquisition expanded their connectomes via multi-synaptic boutons without altering the numbers and spatial arrangements of individual axonal terminals and dendritic spines. This expansion was driven by presynaptic activity elicited by specific negative valence stimuli, regardless of the co-activation state of postsynaptic partners. The rewiring of initial ensembles representing an engram coincided with local, input-specific changes in the shapes and organelle composition of glutamatergic synapses, reflecting their weights and potential for further modifications. Our findings challenge the view that the connectivity among neuronal substrates of memory traces is governed by Hebbian mechanisms, and offer a structural basis for representational drifts.

Engineering the cellulolytic bacterium, Clostridium thermocellum, to co-utilize hemicellulose.

Consolidated bioprocessing (CBP) of lignocellulosic biomass holds promise to realize economic production of second-generation biofuels/chemicals, and Clostridium thermocellum is a leading candidate for CBP due to it being one of the fastest degraders of crystalline cellulose and lignocellulosic biomass. However, CBP by C. thermocellum is approached with co-cultures, because C. thermocellum does not utilize hemicellulose. When compared with a single-species fermentation, the co-culture system introduces unnecessary process complexity that may compromise process robustness. In this study, we engineered C. thermocellum to co-utilize hemicellulose without the need for co-culture. By evolving our previously engineered xylose-utilizing strain in xylose, an evolved clonal isolate (KJC19-9) was obtained and showed improved specific growth rate on xylose by ∼3-fold and displayed comparable growth to a minimally engineered strain grown on the bacteria's naturally preferred substrate, cellobiose. To enable full xylan deconstruction to xylose, we recombinantly expressed three different ß-xylosidase enzymes originating from Thermoanaerobacterium saccharolyticum into KJC19-9 and demonstrated growth on xylan with one of the enzymes. This recombinant strain was capable of co-utilizing cellulose and xylan simultaneously, and we integrated the ß-xylosidase gene into the KJC19-9 genome, creating the KJCBXint strain. The strain, KJC19-9, consumed monomeric xylose but accumulated xylobiose when grown on pretreated corn stover, whereas the final KJCBXint strain showed significantly greater deconstruction of xylan and xylobiose. This is the first reported C. thermocellum strain capable of degrading and assimilating hemicellulose polysaccharide while retaining its cellulolytic capabilities, unlocking significant potential for CBP in advancing the bioeconomy.

Polyphosphate kinase deletion increases laboratory productivity in cyanobacteria.

Identification and manipulation of cellular energy regulation mechanisms may be a strategy to increase productivity in photosynthetic organisms. This work tests the hypothesis that polyphosphate synthesis and degradation play a role in energy management by storing or dissipating energy in the form of ATP. A polyphosphate kinase (ppk) knock-out strain unable to synthesize polyphosphate was generated in the cyanobacterium Synechocystis sp. PCC 6803. This mutant strain demonstrated higher ATP levels and faster growth than the wildtype strain in high-carbon conditions and had a growth defect under multiple stress conditions. In a strain that combined ppk deletion with heterologous expression of ethylene-forming enzyme, higher ethylene productivity was observed than in the wildtype background. These results support the role of polyphosphate synthesis and degradation as an energy regulation mechanism and suggest that such mechanisms may be effective targets in biocontainment design.

Isolation of phosphorus-hyperaccumulating microalgae from revolving algal biofilm (RAB) wastewater treatment systems.

Excess phosphorus (P) in wastewater effluent poses a serious threat to aquatic ecosystems and can spur harmful algal blooms. Revolving algal biofilm (RAB) systems are an emerging technology to recover P from wastewater before discharge into aquatic ecosystems. In RAB systems, a community of microalgae take up and store wastewater P as polyphosphate as they grow in a partially submerged revolving biofilm, which may then be harvested and dried for use as fertilizer in lieu of mined phosphate rock. In this work, we isolated and characterized a total of 101 microalgae strains from active RAB systems across the US Midwest, including 82 green algae, 9 diatoms, and 10 cyanobacteria. Strains were identified by microscopy and 16S/18S ribosomal DNA sequencing, cryopreserved, and screened for elevated P content (as polyphosphate). Seven isolated strains possessed at least 50% more polyphosphate by cell dry weight than a microalgae consortium from a RAB system, with the top strain accumulating nearly threefold more polyphosphate. These top P-hyperaccumulating strains include the green alga Chlamydomonas pulvinata TCF-48 g and the diatoms Eolimna minima TCF-3d and Craticula molestiformis TCF-8d, possessing 11.4, 12.7, and 14.0% polyphosphate by cell dry weight, respectively. As a preliminary test of strain application for recovering P, Chlamydomonas pulvinata TCF-48 g was reinoculated into a bench-scale RAB system containing Bold basal medium. The strain successfully recolonized the system and recovered twofold more P from the medium than a microalgae consortium from a RAB system treating municipal wastewater. These isolated P-hyperaccumulating microalgae may have broad applications in resource recovery from various waste streams, including improving P removal from wastewater.

Large language models encode clinical knowledge.

Large language models (LLMs) have demonstrated impressive capabilities, but the bar for clinical applications is high. Attempts to assess the clinical knowledge of models typically rely on automated evaluations based on limited benchmarks. Here, to address these limitations, we present MultiMedQA, a benchmark combining six existing medical question answering datasets spanning professional medicine, research and consumer queries and a new dataset of medical questions searched online, HealthSearchQA. We propose a human evaluation framework for model answers along multiple axes including factuality, comprehension, reasoning, possible harm and bias. In addition, we evaluate Pathways Language Model1 (PaLM, a 540-billion parameter LLM) and its instruction-tuned variant, Flan-PaLM2 on MultiMedQA. Using a combination of prompting strategies, Flan-PaLM achieves state-of-the-art accuracy on every MultiMedQA multiple-choice dataset (MedQA3, MedMCQA4, PubMedQA5 and Measuring Massive Multitask Language Understanding (MMLU) clinical topics6), including 67.6% accuracy on MedQA (US Medical Licensing Exam-style questions), surpassing the prior state of the art by more than 17%. However, human evaluation reveals key gaps. To resolve this, we introduce instruction prompt tuning, a parameter-efficient approach for aligning LLMs to new domains using a few exemplars. The resulting model, Med-PaLM, performs encouragingly, but remains inferior to clinicians. We show that comprehension, knowledge recall and reasoning improve with model scale and instruction prompt tuning, suggesting the potential utility of LLMs in medicine. Our human evaluations reveal limitations of today's models, reinforcing the importance of both evaluation frameworks and method development in creating safe, helpful LLMs for clinical applications.

Manipulation of glycogen and sucrose synthesis increases photosynthetic productivity in cyanobacteria.

Photosynthetic productivity is limited by low energy conversion efficiency in naturally evolved photosynthetic organisms, via multiple mechanisms that are not fully understood. Here we show evidence that extends recent findings that cyanobacteria use "futile" cycles in the synthesis and degradation of carbon compounds to dissipate ATP. Reduction of the glycogen cycle or the sucrose cycle in the model cyanobacterium Synechocystis 6803 led to redirection of cellular energy toward faster growth under simulated outdoor light conditions in photobioreactors that was accompanied by higher energy charge [concentration ratio of ATP/(ATP + ADP)]. Such manipulation of energy metabolism may have potential in engineering microalgal chassis cells to increase productivity of biomass or target metabolites.

Development of a Machine Learning Model for Sonographic Assessment of Gestational Age.

Importance: Fetal ultrasonography is essential for confirmation of gestational age (GA), and accurate GA assessment is important for providing appropriate care throughout pregnancy and for identifying complications, including fetal growth disorders. Derivation of GA from manual fetal biometry measurements (ie, head, abdomen, and femur) is operator dependent and time-consuming. Objective: To develop artificial intelligence (AI) models to estimate GA with higher accuracy and reliability, leveraging standard biometry images and fly-to ultrasonography videos. Design, Setting, and Participants: To improve GA estimates, this diagnostic study used AI to interpret standard plane ultrasonography images and fly-to ultrasonography videos, which are 5- to 10-second videos that can be automatically recorded as part of the standard of care before the still image is captured. Three AI models were developed and validated: (1) an image model using standard plane images, (2) a video model using fly-to videos, and (3) an ensemble model (combining both image and video models). The models were trained and evaluated on data from the Fetal Age Machine Learning Initiative (FAMLI) cohort, which included participants from 2 study sites at Chapel Hill, North Carolina (US), and Lusaka, Zambia. Participants were eligible to be part of this study if they received routine antenatal care at 1 of these sites, were aged 18 years or older, had a viable intrauterine singleton pregnancy, and could provide written consent. They were not eligible if they had known uterine or fetal abnormality, or had any other conditions that would make participation unsafe or complicate interpretation. Data analysis was performed from January to July 2022. Main Outcomes and Measures: The primary analysis outcome for GA was the mean difference in absolute error between the GA model estimate and the clinical standard estimate, with the ground truth GA extrapolated from the initial GA estimated at an initial examination. Results: Of the total cohort of 3842 participants, data were calculated for a test set of 404 participants with a mean (SD) age of 28.8 (5.6) years at enrollment. All models were statistically superior to standard fetal biometry-based GA estimates derived from images captured by expert sonographers. The ensemble model had the lowest mean absolute error compared with the clinical standard fetal biometry (mean [SD] difference, -1.51 [3.96] days; 95% CI, -1.90 to -1.10 days). All 3 models outperformed standard biometry by a more substantial margin on fetuses that were predicted to be small for their GA. Conclusions and Relevance: These findings suggest that AI models have the potential to empower trained operators to estimate GA with higher accuracy.

Deep Learning Detection of Active Pulmonary Tuberculosis at Chest Radiography Matched the Clinical Performance of Radiologists.

Background The World Health Organization (WHO) recommends chest radiography to facilitate tuberculosis (TB) screening. However, chest radiograph interpretation expertise remains limited in many regions. Purpose To develop a deep learning system (DLS) to detect active pulmonary TB on chest radiographs and compare its performance to that of radiologists. Materials and Methods A DLS was trained and tested using retrospective chest radiographs (acquired between 1996 and 2020) from 10 countries. To improve generalization, large-scale chest radiograph pretraining, attention pooling, and semisupervised learning ("noisy-student") were incorporated. The DLS was evaluated in a four-country test set (China, India, the United States, and Zambia) and in a mining population in South Africa, with positive TB confirmed with microbiological tests or nucleic acid amplification testing (NAAT). The performance of the DLS was compared with that of 14 radiologists. The authors studied the efficacy of the DLS compared with that of nine radiologists using the Obuchowski-Rockette-Hillis procedure. Given WHO targets of 90% sensitivity and 70% specificity, the operating point of the DLS (0.45) was prespecified to favor sensitivity. Results A total of 165 754 images in 22 284 subjects (mean age, 45 years; 21% female) were used for model development and testing. In the four-country test set (1236 subjects, 17% with active TB), the receiver operating characteristic (ROC) curve of the DLS was higher than those for all nine India-based radiologists, with an area under the ROC curve of 0.89 (95% CI: 0.87, 0.91). Compared with these radiologists, at the prespecified operating point, the DLS sensitivity was higher (88% vs 75%, P < .001) and specificity was noninferior (79% vs 84%, P = .004). Trends were similar within other patient subgroups, in the South Africa data set, and across various TB-specific chest radiograph findings. In simulations, the use of the DLS to identify likely TB-positive chest radiographs for NAAT confirmation reduced the cost by 40%-80% per TB-positive patient detected. Conclusion A deep learning method was found to be noninferior to radiologists for the determination of active tuberculosis on digital chest radiographs. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by van Ginneken in this issue.

A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.

Background: Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings. Methods: Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial intelligence (AI) models that used blind sweeps to predict gestational age (GA) and fetal malpresentation. AI GA estimates and standard fetal biometry estimates were compared to a previously established ground truth, and evaluated for difference in absolute error. Fetal malpresentation (non-cephalic vs cephalic) was compared to sonographer assessment. On-device AI model run-times were benchmarked on Android mobile phones. Results: Here we show that GA estimation accuracy of the AI model is non-inferior to standard fetal biometry estimates (error difference -1.4 ± 4.5 days, 95% CI -1.8, -0.9, n = 406). Non-inferiority is maintained when blind sweeps are acquired by novice operators performing only two of six sweep motion types. Fetal malpresentation AUC-ROC is 0.977 (95% CI, 0.949, 1.00, n = 613), sonographers and novices have similar AUC-ROC. Software run-times on mobile phones for both diagnostic models are less than 3 s after completion of a sweep. Conclusions: The gestational age model is non-inferior to the clinical standard and the fetal malpresentation model has high AUC-ROCs across operators and devices. Our AI models are able to run on-device, without internet connectivity, and provide feedback scores to assist in upleveling the capabilities of lightly trained ultrasound operators in low resource settings.

Acetogenic production of 3-Hydroxybutyrate using a native 3-Hydroxybutyryl-CoA Dehydrogenase.

3-Hydroxybutyrate (3HB) is a product of interest as it is a precursor to the commercially produced bioplastic polyhydroxybutyrate. It can also serve as a platform for fine chemicals, medicines, and biofuels, making it a value-added product and feedstock. Acetogens non-photosynthetically fix CO2 into acetyl-CoA and have been previously engineered to convert acetyl-CoA into 3HB. However, as acetogen metabolism is poorly understood, those engineering efforts have had varying levels of success. 3HB, using acetyl-CoA as a precursor, can be synthesized by a variety of different pathways. Here we systematically compare various pathways to produce 3HB in acetogens and discover a native (S)-3-hydroxybutyryl-CoA dehydrogenase, hbd2, responsible for endogenous 3HB production. In conjunction with the heterologous thiolase atoB and CoA transferase ctfAB, hbd2 overexpression improves yields of 3HB on both sugar and syngas (CO/H2/CO2), outperforming the other tested pathways. These results uncovered a previously unknown 3HB production pathway, inform data from prior metabolic engineering efforts, and have implications for future physiological and biotechnological anaerobic research.

An evaluation of Internet searches as a marker of trends in population mental health in the US.

The absence of continuous, real-time mental health assessment has made it challenging to quantify the impacts of the COVID-19 pandemic on population mental health. We examined publicly available, anonymized, aggregated data on weekly trends in Google searches related to anxiety, depression, and suicidal ideation from 2018 to 2020 in the US. We correlated these trends with (1) emergency department (ED) visits for mental health problems and suicide attempts, and (2) surveys of self-reported symptoms of anxiety, depression, and mental health care use. Search queries related to anxiety, depression, and suicidal ideation decreased sharply around March 2020, returning to pre-pandemic levels by summer 2020. Searches related to depression were correlated with the proportion of individuals reporting receiving therapy (r = 0.73), taking medication (r = 0.62) and having unmet mental healthcare needs (r = 0.57) on US Census Household Pulse Survey and modestly correlated with rates of ED visits for mental health conditions. Results were similar when considering instead searches for anxiety. Searches for suicidal ideation did not correlate with external variables. These results suggest aggregated data on Internet searches can provide timely and continuous insights into population mental health and complement other existing tools in this domain.

An uncharacteristically low-potential flavin governs the energy landscape of electron bifurcation.

Electron bifurcation, an energy-conserving process utilized extensively throughout all domains of life, represents an elegant means of generating high-energy products from substrates with less reducing potential. The coordinated coupling of exergonic and endergonic reactions has been shown to operate over an electrochemical potential of ∼1.3 V through the activity of a unique flavin cofactor in the enzyme NADH-dependent ferredoxin-NADP+ oxidoreductase I. The inferred energy landscape has features unprecedented in biochemistry and presents novel energetic challenges, the most intriguing being a large thermodynamically uphill step for the first electron transfer of the bifurcation reaction. However, ambiguities in the energy landscape at the bifurcating site deriving from overlapping flavin spectral signatures have impeded a comprehensive understanding of the specific mechanistic contributions afforded by thermodynamic and kinetic factors. Here, we elucidate an uncharacteristically low two-electron potential of the bifurcating flavin, resolving the energetic challenge of the first bifurcation event.

Genome-Wide Transcription Factor DNA Binding Sites and Gene Regulatory Networks in Clostridium thermocellum.

Clostridium thermocellum is a thermophilic bacterium recognized for its natural ability to effectively deconstruct cellulosic biomass. While there is a large body of studies on the genetic engineering of this bacterium and its physiology to-date, there is limited knowledge in the transcriptional regulation in this organism and thermophilic bacteria in general. The study herein is the first report of a large-scale application of DNA-affinity purification sequencing (DAP-seq) to transcription factors (TFs) from a bacterium. We applied DAP-seq to > 90 TFs in C. thermocellum and detected genome-wide binding sites for 11 of them. We then compiled and aligned DNA binding sequences from these TFs to deduce the primary DNA-binding sequence motifs for each TF. These binding motifs are further validated with electrophoretic mobility shift assay (EMSA) and are used to identify individual TFs' regulatory targets in C. thermocellum. Our results led to the discovery of novel, uncharacterized TFs as well as homologues of previously studied TFs including RexA-, LexA-, and LacI-type TFs. We then used these data to reconstruct gene regulatory networks for the 11 TFs individually, which resulted in a global network encompassing the TFs with some interconnections. As gene regulation governs and constrains how bacteria behave, our findings shed light on the roles of TFs delineated by their regulons, and potentially provides a means to enable rational, advanced genetic engineering of C. thermocellum and other organisms alike toward a desired phenotype.

Early social distancing policies in Europe, changes in mobility & COVID-19 case trajectories: Insights from Spring 2020.

BACKGROUND: Social distancing have been widely used to mitigate community spread of SARS-CoV-2. We sought to quantify the impact of COVID-19 social distancing policies across 27 European counties in spring 2020 on population mobility and the subsequent trajectory of disease. METHODS: We obtained data on national social distancing policies from the Oxford COVID-19 Government Response Tracker and aggregated and anonymized mobility data from Google. We used a pre-post comparison and two linear mixed-effects models to first assess the relationship between implementation of national policies and observed changes in mobility, and then to assess the relationship between changes in mobility and rates of COVID-19 infections in subsequent weeks. RESULTS: Compared to a pre-COVID baseline, Spain saw the largest decrease in aggregate population mobility (~70%), as measured by the time spent away from residence, while Sweden saw the smallest decrease (~20%). The largest declines in mobility were associated with mandatory stay-at-home orders, followed by mandatory workplace closures, school closures, and non-mandatory workplace closures. While mandatory shelter-in-place orders were associated with 16.7% less mobility (95% CI: -23.7% to -9.7%), non-mandatory orders were only associated with an 8.4% decrease (95% CI: -14.9% to -1.8%). Large-gathering bans were associated with the smallest change in mobility compared with other policy types. Changes in mobility were in turn associated with changes in COVID-19 case growth. For example, a 10% decrease in time spent away from places of residence was associated with 11.8% (95% CI: 3.8%, 19.1%) fewer new COVID-19 cases. DISCUSSION: This comprehensive evaluation across Europe suggests that mandatory stay-at-home orders and workplace closures had the largest impacts on population mobility and subsequent COVID-19 cases at the onset of the pandemic. With a better understanding of policies' relative performance, countries can more effectively invest in, and target, early nonpharmacological interventions.

Impacts of social distancing policies on mobility and COVID-19 case growth in the US.

Social distancing remains an important strategy to combat the COVID-19 pandemic in the United States. However, the impacts of specific state-level policies on mobility and subsequent COVID-19 case trajectories have not been completely quantified. Using anonymized and aggregated mobility data from opted-in Google users, we found that state-level emergency declarations resulted in a 9.9% reduction in time spent away from places of residence. Implementation of one or more social distancing policies resulted in an additional 24.5% reduction in mobility the following week, and subsequent shelter-in-place mandates yielded an additional 29.0% reduction. Decreases in mobility were associated with substantial reductions in case growth two to four weeks later. For example, a 10% reduction in mobility was associated with a 17.5% reduction in case growth two weeks later. Given the continued reliance on social distancing policies to limit the spread of COVID-19, these results may be helpful to public health officials trying to balance infection control with the economic and social consequences of these policies.

Deep learning-enabled medical computer vision.

A decade of unprecedented progress in artificial intelligence (AI) has demonstrated the potential for many fields-including medicine-to benefit from the insights that AI techniques can extract from data. Here we survey recent progress in the development of modern computer vision techniques-powered by deep learning-for medical applications, focusing on medical imaging, medical video, and clinical deployment. We start by briefly summarizing a decade of progress in convolutional neural networks, including the vision tasks they enable, in the context of healthcare. Next, we discuss several example medical imaging applications that stand to benefit-including cardiology, pathology, dermatology, ophthalmology-and propose new avenues for continued work. We then expand into general medical video, highlighting ways in which clinical workflows can integrate computer vision to enhance care. Finally, we discuss the challenges and hurdles required for real-world clinical deployment of these technologies.

Transcriptomic analysis of a Clostridium thermocellum strain engineered to utilize xylose: responses to xylose versus cellobiose feeding.

Clostridium (Ruminiclostridium) thermocellum is recognized for its ability to ferment cellulosic biomass directly, but it cannot naturally grow on xylose. Recently, C. thermocellum (KJC335) was engineered to utilize xylose through expressing a heterologous xylose catabolizing pathway. Here, we compared KJC335's transcriptomic responses to xylose versus cellobiose as the primary carbon source and assessed how the bacteria adapted to utilize xylose. Our analyses revealed 417 differentially expressed genes (DEGs) with log2 fold change (FC) >|1| and 106 highly DEGs (log2 FC >|2|). Among the DEGs, two putative sugar transporters, cbpC and cbpD, were up-regulated, suggesting their contribution to xylose transport and assimilation. Moreover, the up-regulation of specific transketolase genes (tktAB) suggests the importance of this enzyme for xylose metabolism. Results also showed remarkable up-regulation of chemotaxis and motility associated genes responding to xylose feeding, as well as widely varying gene expression in those encoding cellulosomal enzymes. For the down-regulated genes, several were categorized in gene ontology terms oxidation-reduction processes, ATP binding and ATPase activity, and integral components of the membrane. This study informs potentially critical, enabling mechanisms to realize the conceptually attractive Next-Generation Consolidated BioProcessing approach where a single species is sufficient for the co-fermentation of cellulose and hemicellulose.

Body condition and poxvirus infection predict circulating glucose levels in a colorful songbird that inhabits urban and rural environments.

There is widespread contemporary interest in causes and consequences of blood glucose status in humans (e.g., links to diabetes and cardiovascular disease), but we know comparatively less about what underlies variation in glucose levels of wild animals. Several environmental factors, including diet, disease status, and habitat quality, may regulate glucose circulation, and we are in need of work that assesses many organismal traits simultaneously to understand the plasticity and predictability of glucose levels in ecological and evolutionary contexts. Here, we measured circulating glucose levels in a species of passerine bird (the house finch, Haemorhous mexicanus) that has served as a valuable model for research on sexual selection, disease, and urban behavioral ecology, as these animals display sexually dichromatic ornamental coloration, harbor many infectious diseases (e.g., poxvirus, coccidiosis, mycoplasmal conjunctivitis), and reside in both natural habitats and cities. We tested the effects of sex, habitat type, body condition, coccidiosis and poxvirus infections, and expression of carotenoid plumage coloration on blood glucose concentrations and found that the body condition and poxvirus infection significantly predicted circulating glucose levels. Specifically, birds with higher blood glucose levels had higher body condition scores and were infected with poxvirus. This result is consistent with biomedical, domesticated-animal, and wildlife-rehabilitation findings, and the premise that glucose elevation is a physiological response to or indicator of infection and relative body weight. The fact that we failed to find links between glucose and our other measurements suggests that blood glucose levels can reveal some but not all aspects of organismal or environmental quality.

Customization scenarios for de-identification of clinical notes.

BACKGROUND: Automated machine-learning systems are able to de-identify electronic medical records, including free-text clinical notes. Use of such systems would greatly boost the amount of data available to researchers, yet their deployment has been limited due to uncertainty about their performance when applied to new datasets. OBJECTIVE: We present practical options for clinical note de-identification, assessing performance of machine learning systems ranging from off-the-shelf to fully customized. METHODS: We implement a state-of-the-art machine learning de-identification system, training and testing on pairs of datasets that match the deployment scenarios. We use clinical notes from two i2b2 competition corpora, the Physionet Gold Standard corpus, and parts of the MIMIC-III dataset. RESULTS: Fully customized systems remove 97-99% of personally identifying information. Performance of off-the-shelf systems varies by dataset, with performance mostly above 90%. Providing a small labeled dataset or large unlabeled dataset allows for fine-tuning that improves performance over off-the-shelf systems. CONCLUSION: Health organizations should be aware of the levels of customization available when selecting a de-identification deployment solution, in order to choose the one that best matches their resources and target performance level.