Disease diagnostics using machine learning of immune receptors.

Clinical diagnosis typically incorporates physical examination, patient history, and various laboratory tests and imaging studies, but makes limited use of the human system's own record of antigen exposures encoded by receptors on B cells and T cells. We analyzed immune receptor datasets from 593 individuals to develop MAchine Learning for Immunological Diagnosis (Mal-ID) , an interpretive framework to screen for multiple illnesses simultaneously or precisely test for one condition. This approach detects specific infections, autoimmune disorders, vaccine responses, and disease severity differences. Human-interpretable features of the model recapitulate known immune responses to SARS-CoV-2, Influenza, and HIV, highlight antigen-specific receptors, and reveal distinct characteristics of Systemic Lupus Erythematosus and Type-1 Diabetes autoreactivity. This analysis framework has broad potential for scientific and clinical interpretation of human immune responses.

Re-visiting humoral constitutive antibacterial heterogeneity in bloodstream infections.

Although cellular immunity has garnered much attention in the era of single-cell technologies, humoral innate immunity has receded in priority due to its presumed limited roles. Hence, despite the long-recognised bactericidal activity of serum-a functional characteristic of constitutive humoral immunity-much remains unclear regarding mechanisms underlying its inter-individual heterogeneity and clinical implications in bloodstream infections. Recent work suggests that the immediate antimicrobial effect of humoral innate immunity contributes to suppression of the excessive inflammatory responses to infection by reducing the amount of pathogen-associated molecular patterns. In this Personal View, we propose the need to re-explore factors underlying the inter-individual heterogeneity in serum antibacterial competence as a new approach to better understand humoral innate immunity and revisit the clinical use of measuring serum antibacterial activity in the management of bacterial bloodstream infections. Given the current emphasis on subtyping sepsis, a serum bactericidal assay might prove useful in defining a distinct sepsis endotype, to enable more personalised management.

Comparative study of encoded and alignment-based methods for virus taxonomy classification.

The emergence of viruses and their variants has made virus taxonomy more important than ever before in controlling the spread of diseases. The creation of efficient treatments and cures that target particular virus properties can be aided by understanding virus taxonomy. Alignment-based methods are commonly used for this task, but are computationally expensive and time-consuming, especially when dealing with large datasets or when detecting new virus variants is time sensitive. An alternative approach, the encoded method, has been developed that does not require prior sequence alignment and provides faster results. However, each encoded method has its own claimed accuracy. Therefore, careful evaluation and comparison of the performance of different encoded methods are essential to identify the most accurate and reliable approach for virus taxonomy classification. This study aims to address this issue by providing a comprehensive and comparative analysis of the potential of encoded methods for virus classification and phylogenetics. We compared the vectors generated for each encoded method using distance metrics to determine their similarity to alignment-based methods. The results and their validation show that K-merNV followed by CgrDft encoded methods, perform similarly to state-of-the-art multi-sequence alignment methods. This is the first study to incorporate and compare encoded methods that will facilitate future research in making more informed decisions regarding selection of a suitable method for virus taxonomy.

Neutrophil Extracellular Traps have DNAzyme activity that drives bactericidal potential.

The mechanisms of bacterial killing by neutrophil extracellular traps (NETs) are unclear. DNA, the largest component of NETs is believed to merely be a scaffold with minimal antimicrobial activity through the charge of the backbone. Here, we report that NETs DNA is beyond a scaffold and produces hydroxyl free radicals through the spatially concentrated G-quadruplex/hemin DNAzyme complexes, driving bactericidal effects. Immunofluorescence staining showed colocalization of G-quadruplex and hemin in extruded NETs DNA, and Amplex UltraRed assay portrayed its peroxidase activity. Proximity labeling of bacteria revealed localized concentration of radicals resulting from NETs bacterial trapping. Ex vivo bactericidal assays revealed that G-quadruplex/hemin DNAzyme is the primary driver of bactericidal activity in NETs. NETs are DNAzymes that may have important biological consequences.

Study of ß1-transferrin and ß2-transferrin using microprobe-capture in-emitter elution and high-resolution mass spectrometry.

Cerebrospinal fluid (CSF) leak can be diagnosed in clinical laboratories by detecting a diagnostic marker ß2-transferrin (ß2-Tf) in secretion samples. ß2-Tf and the typical transferrin (Tf) proteoform in serum, ß1-transferrin (ß1-Tf), are Tf glycoforms. An innovative affinity capture technique for sample preparation, called microprobe-capture in-emitter elution (MPIE), was incorporated with high-resolution mass spectrometry (HR-MS) to study the Tf glycoforms and the primary structures of ß1-Tf and ß2-Tf. To implement MPIE, an analyte is first captured on the surface of a microprobe, and subsequently eluted from the microprobe inside an electrospray emitter. The capture process is monitored in real-time via next-generation biolayer interferometry (BLI). When electrospray is established from the emitter to a mass spectrometer, the analyte is immediately ionized via electrospray ionization (ESI) for HR-MS analysis. Serum, CSF, and secretion samples were analyzed using MPIE-ESI-MS. Based on the MPIE-ESI-MS results, the primary structures of ß1-Tf and ß2-Tf were elucidated. As Tf glycoforms, ß1-Tf and ß2-Tf share the amino acid sequence but contain varying N-glycans: (1) ß1-Tf, the major serum-type Tf, has two G2S2 N-glycans on Asn413 and Asn611; and (2) ß2-Tf, the major brain-type Tf, has an M5 N-glycan on Asn413 and a G0FB N-glycan on Asn611. The resolving power of the innovative MPIE-ESI-MS method was demonstrated in the study of ß2-Tf as well as ß1-Tf. Knowing the N-glycan structures on ß2-Tf allows for the design of more novel test methods for ß2-Tf in the future.

Development and external validation of a pretrained deep learning model for the prediction of non-accidental trauma.

Non-accidental trauma (NAT) is deadly and difficult to predict. Transformer models pretrained on large datasets have recently produced state of the art performance on diverse prediction tasks, but the optimal pretraining strategies for diagnostic predictions are not known. Here we report the development and external validation of Pretrained and Adapted BERT for Longitudinal Outcomes (PABLO), a transformer-based deep learning model with multitask clinical pretraining, to identify patients who will receive a diagnosis of NAT in the next year. We develop a clinical interface to visualize patient trajectories, model predictions, and individual risk factors. In two comprehensive statewide databases, approximately 1% of patients experience NAT within one year of prediction. PABLO predicts NAT events with area under the receiver operating characteristic curve (AUROC) of 0.844 (95% CI 0.838-0.851) in the California test set, and 0.849 (95% CI 0.846-0.851) on external validation in Florida, outperforming comparator models. Multitask pretraining significantly improves model performance. Attribution analysis shows substance use, psychiatric, and injury diagnoses, in the context of age and racial demographics, as influential predictors of NAT. As a clinical decision support system, PABLO can identify high-risk patients and patient-specific risk factors, which can be used to target secondary screening and preventive interventions at the point-of-care.

Phage diversity in cell-free DNA identifies bacterial pathogens in human sepsis cases.

Bacteriophages, viruses that infect bacteria, have great specificity for their bacterial hosts at the strain and species level. However, the relationship between the phageome and associated bacterial population dynamics is unclear. Here we generated a computational pipeline to identify sequences associated with bacteriophages and their bacterial hosts in cell-free DNA from plasma samples. Analysis of two independent cohorts, including a Stanford Cohort of 61 septic patients and 10 controls and the SeqStudy cohort of 224 septic patients and 167 controls, reveals a circulating phageome in the plasma of all sampled individuals. Moreover, infection is associated with overrepresentation of pathogen-specific phages, allowing for identification of bacterial pathogens. We find that information on phage diversity enables identification of the bacteria that produced these phages, including pathovariant strains of Escherichia coli. Phage sequences can likewise be used to distinguish between closely related bacterial species such as Staphylococcus aureus, a frequent pathogen, and coagulase-negative Staphylococcus, a frequent contaminant. Phage cell-free DNA may have utility in studying bacterial infections.

Rapid Molecular Phenotypic Antimicrobial Susceptibility Test for Neisseria gonorrhoeae Based on Propidium Monoazide Viability PCR.

Neisseria gonorrhoeae (NG) is an urgent threat to antimicrobial resistance (AMR) worldwide. NG has acquired rapid resistance to all previously recommended treatments, leaving ceftriaxone monotherapy as the first and last line of therapy for uncomplicated NG. The ability to rapidly determine susceptibility, which is currently nonexistent for NG, has been proposed as a strategy to preserve ceftriaxone by using alternative treatments. Herein, we used a DNA-intercalating dye in combination with NG-specific primers/probes to generate qPCR cycle threshold (Ct) values at different concentrations of 2 NG-relevant antimicrobials. Our proof-of-concept dual-antimicrobial logistic regression model based on the differential Ct measurements achieved an AUC of 0.93 with a categorical agreement for the susceptibility of 84.6%. When surveying the performance against each antimicrobial separately, the model predicted 90 and 75% susceptible and resistant strains, respectively, to ceftriaxone and 66.7 and 83.3% susceptible and resistant strains, respectively, to ciprofloxacin. We further validated the model against the individual replicates and determined the accuracy of the model in classifying susceptibility agnostic of the inoculum size. We demonstrated a novel PCR-based approach to determine phenotypic ciprofloxacin and ceftriaxone susceptibility information for NG with reasonable accuracy within 30 min, a significant improvement compared to the conventional method which could take multiple days.

Rapid molecular phenotypic antimicrobial susceptibility test for Neisseria gonorrhoeae based on propidium monoazide viability PCR.

Neisseria gonorrhoeae (NG) is an urgent threat to antimicrobial resistance (AMR) worldwide. NG has acquired rapid resistance to all previously recommended treatments leaving ceftriaxone monotherapy as the first and last line of therapy for uncomplicated NG. The ability to rapidly determine susceptibility, which is currently nonexistent for NG, has been proposed as a strategy to preserve ceftriaxone by using alternative treatments. Herein, we used a DNA-intercalating dye in combination with NG-specific primers/probes to generate qPCR cycle threshold (Ct) values at different concentrations of 2 NG-relevant antimicrobials. Our proof of concept dual-antimicrobial logistic regression model based on the differential Ct measurements achieved an AUC of 0.93 with a categorical agreement for susceptibility of 84.6%. When surveying the performance against each antimicrobial separately, the model predicted 90% and 75% susceptible and resistant strains respectively to ceftriaxone and 66.7% and 83.3% susceptible and resistant strains respectively to ciprofloxacin. We further validated the model against the individual replicates and determined the accuracy of the model in classifying susceptibility agnostic of the inoculum size. We demonstrated a novel PCR-based approach to determine phenotypic ciprofloxacin and ceftriaxone susceptibility information for NG with reasonable accuracy in under 30 min, a significant improvement compared to the conventional method which takes 3 days.

Microprobe-Capture In-Emitter Elution: An Affinity Capture Technique to Directly Couple a Label-Free Optical Sensing Technology with Mass Spectrometry for Protein Analysis.

Affinity capture of an analyte by a capture agent is one of the most effective sample preparation approaches in mass spectrometry (MS), especially top-down MS. We describe a new affinity capture technique for protein targets, called microprobe-capture in-emitter elution (MPIE), which can directly couple a label-free optical sensing technology (next-generation biolayer interferometry, BLI) with MS. To implement MPIE, an analyte is first captured on the surface of a microprobe and subsequently eluted from the microprobe inside an electrospray emitter. The capture process is monitored in real-time via BLI. When electrospray is established from the emitter to a mass spectrometer, the analyte is immediately ionized via electrospray ionization (ESI) for MS analysis. By this means, BLI and MS are directly coupled in the form of MPIE-ESI-MS. The performance of MPIE-ESI-MS was demonstrated by the analysis of ß-amyloid 1-40 and transferrin using both standard samples and human specimens. In comparison to conventional affinity capture techniques such as bead-based immunoprecipitation, MPIE innovates the affinity capture methodology by introducing real-time process monitoring and providing binding characteristics of analytes, offering more information-rich experiment results. Thus, MPIE is a valuable addition to the top-down MS sample preparation toolbox, and MPIE-ESI-MS can be useful for identification and characterization of targets of interest.

COVID-19 Vaccine: A Common Suspect but Rare Culprit in Drug Rash With Eosinophilia and Systemic Symptoms (DRESS) Syndrome.

Drug rash with eosinophilia and systemic symptoms (DRESS) syndrome is a rare drug reaction that commonly presents with rash, fever, lymphadenopathy, eosinophilia, and multiorgan involvement. We present a case of this syndrome in a 31-year-old male who presented with a diffuse erythematous morbilliform rash with high fever and elevated liver enzymes. Upon history taking, the patient reported acute onset of multiple seizures that required intubation and ICU admission six weeks prior, which started 24 hours after receiving the Johnson and Johnson Janssen coronavirus disease 2019 (COVID-19) vaccine. During that hospitalization, he was given antiseizure medications Keppra (levetiracetam) and Dilantin (phenytoin), which he was eventually discharged home with. During our encounter with the patient, Dermatology was consulted and recommended punch skin biopsy, which revealed spongiotic dermatitis with subcorneal pustules along with superficial perivascular and mixed lymphocytic and neutrophilic infiltrate with dermal edema and rare eosinophils. Given these findings in conjunction with the patient's fever, elevated liver function, and cervical lymphadenopathy, the rash was consistent with DRESS syndrome or a pustular drug eruption likely secondary to phenytoin or levetiracetam. This case was eventually resolved with treatment with oral and topical corticosteroids and close outpatient follow-up with Dermatology. Prompt diagnosis and treatment of DRESS syndrome are therefore critical as the mortality rate can be as high as 10% in the setting of liver failure.

Using a 29-mRNA Host Response Classifier To Detect Bacterial Coinfections and Predict Outcomes in COVID-19 Patients Presenting to the Emergency Department.

Clinicians in the emergency department (ED) face challenges in concurrently assessing patients with suspected COVID-19 infection, detecting bacterial coinfection, and determining illness severity since current practices require separate workflows. Here, we explore the accuracy of the IMX-BVN-3/IMX-SEV-3 29 mRNA host response classifiers in simultaneously detecting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and bacterial coinfections and predicting clinical severity of COVID-19. A total of 161 patients with PCR-confirmed COVID-19 (52.2% female; median age, 50.0 years; 51% hospitalized; 5.6% deaths) were enrolled at the Stanford Hospital ED. RNA was extracted (2.5 mL whole blood in PAXgene blood RNA), and 29 host mRNAs in response to the infection were quantified using Nanostring nCounter. The IMX-BVN-3 classifier identified SARS-CoV-2 infection in 151 patients with a sensitivity of 93.8%. Six of 10 patients undetected by the classifier had positive COVID tests more than 9 days prior to enrollment, and the remaining patients oscillated between positive and negative results in subsequent tests. The classifier also predicted that 6 (3.7%) patients had a bacterial coinfection. Clinical adjudication confirmed that 5/6 (83.3%) of the patients had bacterial infections, i.e., Clostridioides difficile colitis (n = 1), urinary tract infection (n = 1), and clinically diagnosed bacterial infections (n = 3), for a specificity of 99.4%. Two of 101 (2.8%) patients in the IMX-SEV-3 "Low" severity classification and 7/60 (11.7%) in the "Moderate" severity classification died within 30 days of enrollment. IMX-BVN-3/IMX-SEV-3 classifiers accurately identified patients with COVID-19 and bacterial coinfections and predicted patients' risk of death. A point-of-care version of these classifiers, under development, could improve ED patient management, including more accurate treatment decisions and optimized resource utilization. IMPORTANCE We assay the utility of the single-test IMX-BVN-3/IMX-SEV-3 classifiers that require just 2.5 mL of patient blood in concurrently detecting viral and bacterial infections as well as predicting the severity and 30-day outcome from the infection. A point-of-care device, in development, will circumvent the need for blood culturing and drastically reduce the time needed to detect an infection. This will negate the need for empirical use of broad-spectrum antibiotics and allow for antibiotic use stewardship. Additionally, accurate classification of the severity of infection and the prediction of 30-day severe outcomes will allow for appropriate allocation of hospital resources.

Anti-nucleocapsid antibody levels and pulmonary comorbid conditions are linked to post-COVID-19 syndrome.

BACKGROUNDProlonged symptoms after SARS-CoV-2 infection are well documented. However, which factors influence development of long-term symptoms, how symptoms vary across ethnic groups, and whether long-term symptoms correlate with biomarkers are points that remain elusive.METHODSAdult SARS-CoV-2 reverse transcription PCR-positive (RT-PCR-positive) patients were recruited at Stanford from March 2020 to February 2021. Study participants were seen for in-person visits at diagnosis and every 1-3 months for up to 1 year after diagnosis; they completed symptom surveys and underwent blood draws and nasal swab collections at each visit.RESULTSOur cohort (n = 617) ranged from asymptomatic to critical COVID-19 infections. In total, 40% of participants reported at least 1 symptom associated with COVID-19 six months after diagnosis. Median time from diagnosis to first resolution of all symptoms was 44 days; median time from diagnosis to sustained symptom resolution with no recurring symptoms for 1 month or longer was 214 days. Anti-nucleocapsid IgG level in the first week after positive RT-PCR test and history of lung disease were associated with time to sustained symptom resolution. COVID-19 disease severity, ethnicity, age, sex, and remdesivir use did not affect time to sustained symptom resolution.CONCLUSIONWe found that all disease severities had a similar risk of developing post-COVID-19 syndrome in an ethnically diverse population. Comorbid lung disease and lower levels of initial IgG response to SARS-CoV-2 nucleocapsid antigen were associated with longer symptom duration.TRIAL REGISTRATIONClinicalTrials.gov, NCT04373148.FUNDINGNIH UL1TR003142 CTSA grant, NIH U54CA260517 grant, NIEHS R21 ES03304901, Sean N Parker Center for Allergy and Asthma Research at Stanford University, Chan Zuckerberg Biohub, Chan Zuckerberg Initiative, Sunshine Foundation, Crown Foundation, and Parker Foundation.

Biochemical, biophysical, and immunological characterization of respiratory secretions in severe SARS-CoV-2 infections.

Thick, viscous respiratory secretions are a major pathogenic feature of COVID-19, but the composition and physical properties of these secretions are poorly understood. We characterized the composition and rheological properties (i.e., resistance to flow) of respiratory secretions collected from intubated COVID-19 patients. We found the percentages of solids and protein content were greatly elevated in COVID-19 compared with heathy control samples and closely resembled levels seen in cystic fibrosis, a genetic disease known for thick, tenacious respiratory secretions. DNA and hyaluronan (HA) were major components of respiratory secretions in COVID-19 and were likewise abundant in cadaveric lung tissues from these patients. COVID-19 secretions exhibited heterogeneous rheological behaviors, with thicker samples showing increased sensitivity to DNase and hyaluronidase treatment. In histologic sections from these same patients, we observed increased accumulation of HA and the hyaladherin versican but reduced tumor necrosis factor-stimulated gene-6 staining, consistent with the inflammatory nature of these secretions. Finally, we observed diminished type I interferon and enhanced inflammatory cytokines in these secretions. Overall, our studies indicated that increases in HA and DNA in COVID-19 respiratory secretion samples correlated with enhanced inflammatory burden and suggested that DNA and HA may be viable therapeutic targets in COVID-19 infection.

Prevalence of Sensitive Terms in Clinical Notes Using Natural Language Processing Techniques: Observational Study.

BACKGROUND: With the increased sharing of electronic health information as required by the US 21st Century Cures Act, there is an increased risk of breaching patient, parent, or guardian confidentiality. The prevalence of sensitive terms in clinical notes is not known. OBJECTIVE: The aim of this study is to define sensitive terms that represent the documentation of content that may be private and determine the prevalence and characteristics of provider notes that contain sensitive terms. METHODS: Using keyword expansion, we defined a list of 781 sensitive terms. We searched all provider history and physical, progress, consult, and discharge summary notes for patients aged 0-21 years written between January 1, 2019, and December 31, 2019, for a direct string match of sensitive terms. We calculated the prevalence of notes with sensitive terms and characterized clinical encounters and patient characteristics. RESULTS: Sensitive terms were present in notes from every clinical context in all pediatric ages. Terms related to the mental health category were most used overall (254,975/1,338,297, 19.5%), but terms related to substance abuse and reproductive health were most common in patients aged 0-3 years. History and physical notes (19,854/34,771, 57.1%) and ambulatory progress notes (265,302/563,273, 47.1%) were most likely to include sensitive terms. The highest prevalence of notes with sensitive terms was found in pain management (950/1112, 85.4%) and child abuse (1092/1282, 85.2%) clinics. CONCLUSIONS: Notes containing sensitive terms are not limited to adolescent patients, specific note types, or certain specialties. Recognition of sensitive terms across all ages and clinical settings complicates efforts to protect patient and caregiver privacy in the era of information-blocking regulations.

Origin of novel coronavirus causing COVID-19: A computational biology study using artificial intelligence.

Origin of the COVID-19 virus (SARS-CoV-2) has been intensely debated in the scientific community since the first infected cases were detected in December 2019. The disease has caused a global pandemic, leading to deaths of thousands of people across the world and thus finding origin of this novel coronavirus is important in responding and controlling the pandemic. Recent research results suggest that bats or pangolins might be the hosts for SARS-CoV-2 based on comparative studies using its genomic sequences. This paper investigates the SARS-CoV-2 origin by using artificial intelligence (AI)-based unsupervised learning algorithms and raw genomic sequences of the virus. More than 300 genome sequences of COVID-19 infected cases collected from different countries are explored and analysed using unsupervised clustering methods. The results obtained from various AI-enabled experiments using clustering algorithms demonstrate that all examined SARS-CoV-2 genomes belong to a cluster that also contains bat and pangolin coronavirus genomes. This provides evidence strongly supporting scientific hypotheses that bats and pangolins are probable hosts for SARS-CoV-2. At the whole genome analysis level, our findings also indicate that bats are more likely the hosts for the COVID-19 virus than pangolins.

Diagnosis of Bloodstream Infections: An Evolution of Technologies towards Accurate and Rapid Identification and Antibiotic Susceptibility Testing.

Bloodstream infections (BSI) are a leading cause of death worldwide. The lack of timely and reliable diagnostic practices is an ongoing issue for managing BSI. The current gold standard blood culture practice for pathogen identification and antibiotic susceptibility testing is time-consuming. Delayed diagnosis warrants the use of empirical antibiotics, which could lead to poor patient outcomes, and risks the development of antibiotic resistance. Hence, novel techniques that could offer accurate and timely diagnosis and susceptibility testing are urgently needed. This review focuses on BSI and highlights both the progress and shortcomings of its current diagnosis. We surveyed clinical workflows that employ recently approved technologies and showed that, while offering improved sensitivity and selectivity, these techniques are still unable to deliver a timely result. We then discuss a number of emerging technologies that have the potential to shorten the overall turnaround time of BSI diagnosis through direct testing from whole blood-while maintaining, if not improving-the current assay's sensitivity and pathogen coverage. We concluded by providing our assessment of potential future directions for accelerating BSI pathogen identification and the antibiotic susceptibility test. While engineering solutions have enabled faster assay turnaround, further progress is still needed to supplant blood culture practice and guide appropriate antibiotic administration for BSI patients.

Biochemical, Biophysical, and Immunological Characterization of Respiratory Secretions in Severe SARS-CoV-2 (COVID-19) Infections.

Thick, viscous respiratory secretions are a major pathogenic feature of COVID-19 disease, but the composition and physical properties of these secretions are poorly understood. We characterized the composition and rheological properties (i.e. resistance to flow) of respiratory secretions collected from intubated COVID-19 patients. We find the percent solids and protein content are greatly elevated in COVID-19 compared to heathy control samples and closely resemble levels seen in cystic fibrosis, a genetic disease known for thick, tenacious respiratory secretions. DNA and hyaluronan (HA) are major components of respiratory secretions in COVID-19 and are likewise abundant in cadaveric lung tissues from these patients. COVID-19 secretions exhibit heterogeneous rheological behaviors with thicker samples showing increased sensitivity to DNase and hyaluronidase treatment. In histologic sections from these same patients, we observe increased accumulation of HA and the hyaladherin versican but reduced tumor necrosis factorâ€"stimulated gene-6 (TSG6) staining, consistent with the inflammatory nature of these secretions. Finally, we observed diminished type I interferon and enhanced inflammatory cytokines in these secretions. Overall, our studies indicate that increases in HA and DNA in COVID-19 respiratory secretion samples correlate with enhanced inflammatory burden and suggest that DNA and HA may be viable therapeutic targets in COVID-19 infection.

Integrating deep learning and unbiased automated high-content screening to identify complex disease signatures in human fibroblasts.

Drug discovery for diseases such as Parkinson's disease are impeded by the lack of screenable cellular phenotypes. We present an unbiased phenotypic profiling platform that combines automated cell culture, high-content imaging, Cell Painting, and deep learning. We applied this platform to primary fibroblasts from 91 Parkinson's disease patients and matched healthy controls, creating the largest publicly available Cell Painting image dataset to date at 48 terabytes. We use fixed weights from a convolutional deep neural network trained on ImageNet to generate deep embeddings from each image and train machine learning models to detect morphological disease phenotypes. Our platform's robustness and sensitivity allow the detection of individual-specific variation with high fidelity across batches and plate layouts. Lastly, our models confidently separate LRRK2 and sporadic Parkinson's disease lines from healthy controls (receiver operating characteristic area under curve 0.79 (0.08 standard deviation)), supporting the capacity of this platform for complex disease modeling and drug screening applications.

Detection of bacterial co-infections and prediction of fatal outcomes in COVID-19 patients presenting to the emergency department using a 29 mRNA host response classifier.

Objective: Clinicians in the emergency department (ED) face challenges in concurrently assessing patients with suspected COVID-19 infection, detecting bacterial co-infection, and determining illness severity since current practices require separate workflows. Here we explore the accuracy of the IMX-BVN-3/IMX-SEV-3 29 mRNA host response classifiers in simultaneously detecting SARS-CoV-2 infection, bacterial co-infections, and predicting clinical severity of COVID-19. Methods: 161 patients with PCR-confirmed COVID-19 (52.2% female, median age 50.0 years, 51% hospitalized, 5.6% deaths) were enrolled at the Stanford Hospital ED. RNA was extracted (2.5 mL whole blood in PAXgene Blood RNA) and 29 host mRNAs in response to the infection were quantified using Nanostring nCounter. Results: The IMX-BVN-3 classifier identified SARS-CoV-2 infection in 151 patients with a sensitivity of 93.8%. Six of 10 patients undetected by the classifier had positive COVID tests more than 9 days prior to enrolment and the remaining oscillated between positive and negative results in subsequent tests. The classifier also predicted that 6 (3.7%) patients had a bacterial co-infection. Clinical adjudication confirmed that 5/6 (83.3%) of the patients had bacterial infections, i.e. Clostridioides difficile colitis (n=1), urinary tract infection (n=1), and clinically diagnosed bacterial infections (n=3) for a specificity of 99.4%. 2/101 (2.8%) patients in the IMX-SEV-3 Low and 7/60 (11.7%) in the Moderate severity classifications died within thirty days of enrollment. Conclusions: IMX-BVN-3/IMX-SEV-3 classifiers accurately identified patients with COVID-19, bacterial co-infections, and predicted patients’ risk of death. A point-of-care version of these classifiers, under development, could improve ED patient management including more accurate treatment decisions and optimized resource utilization.