Machine learning-based clinical decision support for infection risk prediction.

Background: Healthcare-associated infection (HAI) remains a significant risk for hospitalized patients and a challenging burden for the healthcare system. This study presents a clinical decision support tool that can be used in clinical workflows to proactively engage secondary assessments of pre-symptomatic and at-risk infection patients, thereby enabling earlier diagnosis and treatment. Methods: This study applies machine learning, specifically ensemble-based boosted decision trees, on large retrospective hospital datasets to develop an infection risk score that predicts infection before obvious symptoms present. We extracted a stratified machine learning dataset of 36,782 healthcare-associated infection patients. The model leveraged vital signs, laboratory measurements and demographics to predict HAI before clinical suspicion, defined as the order of a microbiology test or administration of antibiotics. Results: Our best performing infection risk model achieves a cross-validated AUC of 0.88 at 1 h before clinical suspicion and maintains an AUC >0.85 for 48 h before suspicion by aggregating information across demographics and a set of 163 vital signs and laboratory measurements. A second model trained on a reduced feature space comprising demographics and the 36 most frequently measured vital signs and laboratory measurements can still achieve an AUC of 0.86 at 1 h before clinical suspicion. These results compare favorably against using temperature alone and clinical rules such as the quick sequential organ failure assessment (qSOFA) score. Along with the performance results, we also provide an analysis of model interpretability via feature importance rankings. Conclusion: The predictive model aggregates information from multiple physiological parameters such as vital signs and laboratory measurements to provide a continuous risk score of infection that can be deployed in hospitals to provide advance warning of patient deterioration.

A Retrospective Cohort Study of Clinical Factors Associated with Transitions of Care among COVID-19 Patients.

Coronavirus Disease 2019 (COVID-19) is an international health crisis. In this article, we report on patient characteristics associated with care transitions of: 1) hospital admission from the emergency department (ED) and 2) escalation to the intensive care unit (ICU). Analysis of data from the electronic medical record (EMR) was performed for patients with COVID-19 seen in the ED of a large Western U.S. Health System from April to August of 2020, totaling 10,079 encounters. Of these, 5172 resulted in admission as an inpatient within 72 h. Inpatient encounters (n = 6079) were also considered for patients with positive COVID-19 test results, of which 970 resulted in a transfer to the ICU or in-hospital mortality. Laboratory results, vital signs, symptoms, and comorbidities were investigated for each of these care transitions. Different top risk factors were found, but two factors common to hospital admission and ICU transfer were respiratory rate and the need for oxygen support. Comorbidities common to both settings were cerebrovascular disease and congestive heart failure. Regarding laboratory results, the neutrophil-to-lymphocyte ratio was associated with transitions to higher levels of care, along with the ratio of aspartate aminotransferase (AST) to alanine aminotransferase (ALT).

Evolution-informed modeling improves outcome prediction for cancers.

Despite wide applications of high-throughput biotechnologies in cancer research, many biomarkers discovered by exploring large-scale omics data do not provide satisfactory performance when used to predict cancer treatment outcomes. This problem is partly due to the overlooking of functional implications of molecular markers. Here, we present a novel computational method that uses evolutionary conservation as prior knowledge to discover bona fide biomarkers. Evolutionary selection at the molecular level is nature's test on functional consequences of genetic elements. By prioritizing genes that show significant statistical association and high functional impact, our new method reduces the chances of including spurious markers in the predictive model. When applied to predicting therapeutic responses for patients with acute myeloid leukemia and to predicting metastasis for patients with prostate cancers, the new method gave rise to evolution-informed models that enjoyed low complexity and high accuracy. The identified genetic markers also have significant implications in tumor progression and embrace potential drug targets. Because evolutionary conservation can be estimated as a gene-specific, position-specific, or allele-specific parameter on the nucleotide level and on the protein level, this new method can be extended to apply to miscellaneous "omics" data to accelerate biomarker discoveries.

A Crowdsourcing Approach to Developing and Assessing Prediction Algorithms for AML Prognosis.

Acute Myeloid Leukemia (AML) is a fatal hematological cancer. The genetic abnormalities underlying AML are extremely heterogeneous among patients, making prognosis and treatment selection very difficult. While clinical proteomics data has the potential to improve prognosis accuracy, thus far, the quantitative means to do so have yet to be developed. Here we report the results and insights gained from the DREAM 9 Acute Myeloid Prediction Outcome Prediction Challenge (AML-OPC), a crowdsourcing effort designed to promote the development of quantitative methods for AML prognosis prediction. We identify the most accurate and robust models in predicting patient response to therapy, remission duration, and overall survival. We further investigate patient response to therapy, a clinically actionable prediction, and find that patients that are classified as resistant to therapy are harder to predict than responsive patients across the 31 models submitted to the challenge. The top two performing models, which held a high sensitivity to these patients, substantially utilized the proteomics data to make predictions. Using these models, we also identify which signaling proteins were useful in predicting patient therapeutic response.

Endothelial cells decode VEGF-mediated Ca2+ signaling patterns to produce distinct functional responses.

A single extracellular stimulus can promote diverse behaviors among isogenic cells by differentially regulated signaling networks. We examined Ca(2+) signaling in response to VEGF (vascular endothelial growth factor), a growth factor that can stimulate different behaviors in endothelial cells. We found that altering the amount of VEGF signaling in endothelial cells by stimulating them with different VEGF concentrations triggered distinct and mutually exclusive dynamic Ca(2+) signaling responses that correlated with different cellular behaviors. These behaviors were cell proliferation involving the transcription factor NFAT (nuclear factor of activated T cells) and cell migration involving MLCK (myosin light chain kinase). Further analysis suggested that this signal decoding was robust to the noisy nature of the signal input. Using probabilistic modeling, we captured both the stochastic and deterministic aspects of Ca(2+) signal decoding and accurately predicted cell responses in VEGF gradients, which we used to simulate different amounts of VEGF signaling. Ca(2+) signaling patterns associated with proliferation and migration were detected during angiogenesis in developing zebrafish.

Inferring causal molecular networks: empirical assessment through a community-based effort.

It remains unclear whether causal, rather than merely correlational, relationships in molecular networks can be inferred in complex biological settings. Here we describe the HPN-DREAM network inference challenge, which focused on learning causal influences in signaling networks. We used phosphoprotein data from cancer cell lines as well as in silico data from a nonlinear dynamical model. Using the phosphoprotein data, we scored more than 2,000 networks submitted by challenge participants. The networks spanned 32 biological contexts and were scored in terms of causal validity with respect to unseen interventional data. A number of approaches were effective, and incorporating known biology was generally advantageous. Additional sub-challenges considered time-course prediction and visualization. Our results suggest that learning causal relationships may be feasible in complex settings such as disease states. Furthermore, our scoring approach provides a practical way to empirically assess inferred molecular networks in a causal sense.