Phenotype Assignment Research Articles

Evaluation of repeated machine learning-based phenotyping in patients with cardiogenic shock

Abstract Background Cardiogenic shock (CS) is a highly mortal disease with variable clinical manifestations, severity, and in-hospital trajectories. Multiple clinical trials have failed to show an impact of therapies on outcomes, partly because they focus on etiology as a classifier, failing to capture the clinical heterogeneity. A phenotype-based approach to CS allows for a dynamic risk stratification based on disease severity. We have previously presented three CS phenotypes at baseline using machine learning-based algorithms: phenotype I (non-congested), II (cardiorenal), and III (cardiometabolic). However, little is known about the clinical course of these CS phenotypes during hospitalization and whether repeated phenotyping may provide additional insights. Purpose To assess relevance and prognostic impact of repetitive reassessment of CS phenotypes and their trajectories during the first 72h after CS diagnosis. Methods We included all-cause CS patients from the multi-center, multinational Cardiogenic Shock Working Group (CSWG) V3 and V4 registries (Yr. 2022-2023). Phenotypes were assigned by nearest centroid classification based on the centroids from the initial CSWG V1 (Yr. 2019) registry derivation cohort. Missing values required for phenotype assignment were imputed for all patients, but imputed values were only used for phenotype assignment and not reported as outcomes. Results A total of 4308 patients were included, of which 928 had CS due to myocardial infarction (MI-CS) and 2164 had CS due to heart failure (HF-CS). At time of diagnosis/ baseline, 1627 (37.8%), 1832 (42.5%), and 849 (19.7%) of patients were in phenotype I (non-congested), II (cardiorenal) and III (cardiometabolic), respectively. In MI-CS, patients were evenly distributed between the three phenotypes (35.5%, 34.4%, and 30.2%, respectively), whereas in HF-CS, non-congested and cardiorenal CS were more common than cardiometabolic shock (38.8%, 48.9%, and 12.3%, respectively). Within the first 12 hours, most patients in cardiorenal and cardiometabolic CS shifted to non-congested CS (Figure 1). In-hospital mortality in admission phenotype I, II, and III was 21.7%, 32.1%, and 50.1% (p<0.0001 for difference), respectively, in concordance with previous reports. When assessed at different time points, phenotypes II and III were associated with higher odds of mortality than phenotype I at 24h, 48h, and 72h after initial shock diagnosis (p≤0.03 for all, Figure 2). Conclusion Beyond baseline, re-assessment of machine learning-based CS phenotypes within the first 72h after CS diagnosis may provide additional prognostic insights. These data may inform future clinical trial design to tailor interventions to specific CS phenotypes.Figure 1Figure 2

Clinical Phenotyping for Prognosis and Immunotherapy Guidance in Bacterial Sepsis and COVID-19.

It is suggested that sepsis may be classified into four clinical phenotypes, using an algorithm employing 29 admission parameters. We applied a simplified phenotyping algorithm among patients with bacterial sepsis and severe COVID-19 and assessed characteristics and outcomes of the derived phenotypes. Retrospective analysis of data from prospective clinical studies. Greek ICUs and Internal Medicine departments. We analyzed 1498 patients, 620 with bacterial sepsis and 878 with severe COVID-19. We implemented a six-parameter algorithm (creatinine, lactate, aspartate transaminase, bilirubin, C-reactive protein, and international normalized ratio) to classify patients with bacterial sepsis intro previously defined phenotypes. Patients with severe COVID-19, included in two open-label immunotherapy trials were subsequently classified. Heterogeneity of treatment effect of anakinra was assessed. The primary outcome was 28-day mortality. The algorithm validated the presence of the four phenotypes across the cohort of bacterial sepsis and the individual studies included in this cohort. Phenotype α represented younger patients with low risk of death, β was associated with high comorbidity burden, and δ with the highest mortality. Phenotype assignment was independently associated with outcome, even after adjustment for Charlson Comorbidity Index. Phenotype distribution and outcomes in severe COVID-19 followed a similar pattern. A simplified algorithm successfully identified previously derived phenotypes of bacterial sepsis, which were predictive of outcome. This classification may apply to patients with severe COVID-19 with prognostic implications.

Pharmacogenomic insights in psychiatric care: uncovering novel actionability, allele-specific CYP2D6 copy number variation, and phenoconversion in 15,000 patients

Pharmacogenomic testing has emerged as an aid in clinical decision making for psychiatric providers, but more data is needed regarding its utility in clinical practice and potential impact on patient care. In this cross-sectional study, we determined the real-world prevalence of pharmacogenomic actionability in patients receiving psychiatric care. Potential actionability was based on the prevalence of CYP2C19 and CYP2D6 phenotypes, including CYP2D6 allele-specific copy number variations (CNVs). Combined actionability additionally incorporated CYP2D6 phenoconversion and the novel CYP2C-TG haplotype in patients with available medication data. Across 15,000 patients receiving clinical pharmacogenomic testing, 65% had potentially actionable CYP2D6 and CYP2C19 phenotypes, and phenotype assignment was impacted by CYP2D6 allele-specific CNVs in 2% of all patients. Of 4114 patients with medication data, 42% had CYP2D6 phenoconversion from drug interactions and 20% carried a novel CYP2C haplotype potentially altering actionability. A total of 87% had some form of potential actionability from genetic findings and/or phenoconversion. Genetic variation detected via next-generation sequencing led to phenotype reassignment in 22% of individuals overall (2% in CYP2D6 and 20% in CYP2C19). Ultimately, pharmacogenomic testing using next-generation sequencing identified potential actionability in most patients receiving psychiatric care. Early pharmacogenomic testing may provide actionable insights to aid clinicians in drug prescribing to optimize psychiatric care.

Assessment of the phenotypic severity of hemophilia A: using rotational thromboelastometry (ROTEM) and APTT-clot waveform analysis

BackgroundHemophilia A (HA) is an X-linked inherited bleeding disorder caused by reduced factor VIII (FVIII) levels. Approximately 10–15% of patients with severe HA (SHA) do not present with the anticipated bleeding pattern. Here, we assessed the phenotypic severity of hemophilia A using rotational thromboelastometry (ROTEM) and activated partial thromboplastin time-clot waveform analysis (APTT-CWA).MethodsPatients diagnosed with hemophilia A were enrolled. Clinical phenotype assignment was performed according to the published literature, and patients were classified into four phenotypic subgroups. The whole blood sample was first run on ROTEM in INTEM mode using platelet-poor plasma, APTT was run, and the APTT-CWA graph was simultaneously recorded.ResultsA total of 66 patients were recruited for this study. Statistically significant differences were observed between the four phenotypically categorized groups using ROTEM and APTT-CWA. On comparing patients with mild/moderate-to-severe phenotypes (Group II) with SHA without inhibitors (Group IV), no significant difference was found for all parameters of ROTEM or APTT-CWA. The MCF, MA30, MAXV, and Alpha angle values using ROTEM were found to be the lowest in patients with SHA with inhibitors, which helped differentiate them from those with SHA without inhibitors. However, these two groups could not be differentiated using the APTT-CWA parameters.ConclusionROTEM can be used to distinguish patients with SHA with inhibitors from those with SHA without inhibitors using a combination of parameters with high sensitivity and specificity. However, APTT-CWA cannot be used to differentiate these patient groups.

Abstract PO3-28-10: MRI Radiomic phenotypes derived from the ECOG-ACRIN E4112 Trial to assess high-risk ductal carcinoma in situ

Abstract INTRODUCTION Ductal carcinoma in situ (DCIS) is a non-lethal pre-invasive breast cancer that can co-exist with invasive disease. Dynamic contrast-enhanced (DCE) MRI is sensitive for the detection of high-grade DCIS and invasive cancer while Oncotype DCIS Score is a 12-gene assay that can assess recurrence risk. In practice, it is difficult to distinguish low- from high-risk DCIS, which leads to overtreatment for up to half of women diagnosed with DCIS. We hypothesize that radiomic phenotypes of DCIS derived from DCE-MRI data may serve as prognostic markers to improve risk stratification by capturing disease heterogeneity. Here we evaluate the ability of these phenotypes to predict DCIS Score and upstaging of DCIS to invasive disease on wide local excision in a multicenter trial. METHODS Data: DCE-MRI data from the ECOG-ACRIN E4112 trial were retrospectively analyzed. Primary analysis focused on participants with data on disease upstaging (N=295), with secondary analysis in a subset (N=174) of participants with DCIS Scores (dichotomized as >55 and 55) and pure DCIS. Clinical information included patient demographics, lesion morphology on MRI, background parenchymal enhancement, DCIS grade, central necrosis, and hormone receptor status. Data analysis: Radiologist-drawn lesion segmentations and publicly available software, CaPTk, were used to compute 64 radiomic features from first post-contrast images for each participant. Radiomic phenotypes were identified using hierarchical clustering on the extracted features. A Chi-square test was used to evaluate the association between radiomic phenotypes and each outcome. The likelihood ratio test was used to compare two logistic regression models: 1)clinical model using only clinical information as predictors and 2) clinical+phenotypes model using clinical information and phenotype assignment as predictors. Each model was used for the prediction of upstaging to invasive disease and DCIS score. Model performance was evaluated as the 10-fold cross-validated area under the receiver operator characteristic curve (AUC). A p< 0.05 was considered significant. RESULTS A total of 45 (15%) cases upstaged to invasive disease. Two radiomic phenotypes were identified: Phenotype 1 indicated greater lesion signal heterogeneity, while Phenotype 2 indicated lower heterogeneity. Radiomic phenotype was strongly associated with disease upstaging (p=0.0034) – with a higher rate of upstaging for Phenotype 1 – but not with DCIS score (p=0.1174, Table 1). For predicting disease upstaging, the clinical+phenotypes model yielded a higher AUC=0.72 and a significantly better fit to the data (p=0.0022) compared to the clinical model parameterized by clinical information alone (AUC=0.69). For predicting DCIS Score, the clinical+phenotypes model (AUC=0.77) showed similar performance compared to the clinical model (AUC=0.76) and no significant improvement in fit to the data (p=0.2920). CONCLUSION Radiomic phenotypes capturing disease heterogeneity show promise as prognostic predictors for predicting disease upstaging in DCIS compared to clinical information alone and may enable more efficient disease management. We observed that phenotypes did not have independent predictive outcome for DCIS score, suggesting that MRI and DCIS Score offer independent information and could be combined in future models to better predict disease recurrence or progression. Clinical applications of radiomic phenotypes may improve risk stratification and potentially result in decreased overtreatment of women diagnosed with DCIS. Table 1: Association between radiomic phenotypes and DCIS outcomes using a Chi square test. Citation Format: Kalina Slavkova, Ruya Kang, Vivian Belenky, Anum Kazerouni, Debosmita Biswas, Hannah Horng, Rhea Chitalia, Michael Hirano, Jennifer Xiao, Ralph Corsetti, Sarah Javid, Derrick Spell, Antonio Wolff, Joseph Sparano, Seema Khan, Christopher Comstock, Justin Romanoff, Jon Steingrimsson, Constantine Gatsonis, Constance Lehman, Savannah Partridge, Despina Kontos, Habib Rahbar. MRI Radiomic phenotypes derived from the ECOG-ACRIN E4112 Trial to assess high-risk ductal carcinoma in situ [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO3-28-10.

Does psychological distress predict risk of orthopaedic surgery and postoperative opioid prescribing in patients with hip pain? A retrospective study

BackgroundClinicians and public health professionals have allocated resources to curb opioid over-prescription and address psychological needs among patients with musculoskeletal pain. However, associations between psychological distress, risk of surgery, and opioid prescribing among those with hip pathologies remain unclear.MethodsUsing a retrospective cohort study design, we identified patients that were evaluated for hip pain from January 13, 2020 to October 27, 2021. Patients’ surgical histories and postoperative opioid prescriptions were extracted via chart review. Risk of hip surgery within one year of evaluation was analyzed using multivariable logistic regression. Multivariable linear regression was employed to predict average morphine milligram equivalents (MME) per day of opioid prescriptions within the first 30 days after surgery. Candidate predictors included age, gender, race, ethnicity, employment, insurance type, hip function and quality of life on the International Hip Outcome Tool (iHOT-12), and psychological distress phenotype using the OSPRO Yellow Flag (OSPRO-YF) Assessment Tool.ResultsOf the 672 patients, n = 350 (52.1%) underwent orthopaedic surgery for hip pain. In multivariable analysis, younger patients, those with TRICARE/other government insurance, and those with a high psychological distress phenotype had higher odds of surgery. After adding iHOT-12 scores, younger patients and lower iHOT-12 scores were associated with higher odds of surgery, while Black/African American patients had lower odds of surgery. In multivariable analysis of average MME, patients with periacetabular osteotomy (PAO) received opioid prescriptions with significantly higher average MME than those with other procedures, and surgery type was the only significant predictor. Post-hoc analysis excluding PAO found higher average MME for patients undergoing hip arthroscopy (compared to arthroplasty or other non-PAO procedures) and significantly lower average MME for patients with public insurance (Medicare/Medicaid) compared to those with private insurance. Among those only undergoing arthroscopy, older age and having public insurance were associated with opioid prescriptions with lower average MME. Neither iHOT-12 scores nor OSPRO-YF phenotype assignment were significant predictors of postoperative mean MME.ConclusionsPsychological distress characteristics are modifiable targets for rehabilitation programs, but their use as prognostic factors for risk of orthopaedic surgery and opioid prescribing in patients with hip pain appears limited when considered alongside other commonly collected clinical information such as age, insurance, type of surgery pursued, and iHOT-12 scores.

Prevalence of CYP2D6 structural variation in large retrospective study.

CYP2D6 is a highly polymorphic gene with clinically important structural variations. Commonly, only exon 9 is assayed on clinical pharmacogenomics panels, as it allows for accurate functional characterization even in the presence of a CYP2D6::CYP2D7 conversion. However, this method does not capture CYP2D7::CYP2D6 (CYP2D6*13) conversions, possibly leading to inaccurate phenotype assignment. The study's purpose was to determine the frequency of structural variations in CYP2D6 utilizing multiple copy number variation (CNV) assay locations to quantify the potential impact on clinical phenotype classification. A retrospective analysis was conducted of de-identified pharmacogenomics data submitted through the Translational Software, Inc. platform. Samples with CYP2D6 CNV data for exon 9 and at least one additional CNV location (5'UTR, exon 1, intron 2, exon 5 or intron 6) were included. CYP2D7::CYP2D6 and CYP2D6::CYP2D7 conversions were classified according to PharmVar nomenclature. The CYP2D6 copies were capped at four total copies to account for assay limitations in detecting more than four copies. A total of 106,474 samples were included for analysis. CYP2D7::CYP2D6 conversions were present in approximately 2.44% of samples, and 5.84% of samples had CYP2D6::CYP2D7 conversions. Many samples did not have a CYP2D7 conversion detected (91.5%; 97,462/106,474). A full gene deletion was detected in 0.15%, and 5.98% had a duplication or multiplication present. This retrospective study underscores the importance of testing more than one CNV site for CYP2D6 . Over 2% of patients were found to have a CYP2D7::CYP2D6 conversion. This translates into potentially misclassified phenotype classification and incongruent clinical recommendations.

Clinical outcomes of patients in cardiogenic shock phenotypes: data from the cardiogenic shock working group registry

Abstract Background Despite therapeutic innovations in cardiogenic shock (CS), CS patients remain at high in-hospital mortality of 30-60%. Heterogeneity of CS patients has complicated clinical trial design. Recently, three CS phenotypes were identified in Version 1 (V1) of the multi-center national Cardiogenic Shock Working Group (CSWG) registry and two validation cohorts, namely phenotype I: "non-congested", II: "cardiorenal", and III: "cardiometabolic" shock. However, little is known about the clinical course of these CS phenotypes and further validation is required. Purpose (i) To confirm the external applicability of phenotypes in a mixed CS population, (ii) to assess the relevance of CS phenotypes for prognosis and outcomes, (iii) to confirm their compatibility with the SCAI classification system, and (iv) to assess differences in temporary mechanical circulatory support (tMCS) usage among CS phenotypes. Methods We included all-cause CS patients from CSWG V2 registry. Phenotypes were assigned by nearest centroid classification with the centroids from the initial CSWG V1 registry derivation cohort. Analyses were performed in all patients with complete clustering data. A sensitivity analysis in the complete dataset included cluster assignments based on random forest imputation. Results 796 out of 1890 all-cause CS patients had complete data for retrospective phenotype assignment without imputation. Among these, mortality in phenotype I (non-congested), II (cardiorenal) and III (cardiometabolic) was 23%, 41% and 52%, respectively, closely matching the initially reported 21%, 45% and 55% in the V1 derivation cohort. In the total cohort including patients with partially imputed data, mortality was 23%, 30%, and 52%, respectively. Phenotype-related demographic, hemodynamic and metabolic features resembled the initially reported properties (Figure). In the main cohort of 796 patients, 58.8%, 45.7%, and 51.9% in phenotype I, II, and III, received tMCS, respectively (p=0.013). Receiving any tMCS was associated with higher mortality in cardiorenal (OR [95%-CI]: 1.82 [1.16-2.84; p=0.008), but not in non-congested or cardiometabolic CS (1.26 [0.64-2.47], p=0.51 & 1.39 [0.86-2.25], p=0.18, respectively). Patients in cardiometabolic CS but not cardiorenal CS at baseline were more likely to reach SCAI Stage E ("Extremis") during the hospitalization (OR [95% CI]: 5.26 [3.53-7.83], p<0.001 & 1.10 [0.76-1.61], p=0.609, respectively) than patients with non-congested CS at baseline. Admission phenotype II and III as well as admission SCAI stage E were each independently associated with more than twofold increases in odds of mortality compared to phenotype I and baseline SCAI stage C in a multivariable regression (p<0.001, respectively). Conclusion Our findings support the universal applicability of phenotypes in all-cause CS. These data may inform future clinical trial design to tailor interventions to specific CS phenotypes.Figure

Interpretable clinical phenotypes among patients hospitalized with COVID-19 using cluster analysis.

Multiple clinical phenotypes have been proposed for coronavirus disease (COVID-19), but few have used multimodal data. Using clinical and imaging data, we aimed to identify distinct clinical phenotypes in patients admitted with COVID-19 and to assess their clinical outcomes. Our secondary objective was to demonstrate the clinical applicability of this method by developing an interpretable model for phenotype assignment. We analyzed data from 547 patients hospitalized with COVID-19 at a Canadian academic hospital. We processed the data by applying a factor analysis of mixed data (FAMD) and compared four clustering algorithms: k-means, partitioning around medoids (PAM), and divisive and agglomerative hierarchical clustering. We used imaging data and 34 clinical variables collected within the first 24 h of admission to train our algorithm. We conducted a survival analysis to compare the clinical outcomes across phenotypes. With the data split into training and validation sets (75/25 ratio), we developed a decision-tree-based model to facilitate the interpretation and assignment of the observed phenotypes. Agglomerative hierarchical clustering was the most robust algorithm. We identified three clinical phenotypes: 79 patients (14%) in Cluster 1, 275 patients (50%) in Cluster 2, and 203 (37%) in Cluster 3. Cluster 2 and Cluster 3 were both characterized by a low-risk respiratory and inflammatory profile but differed in terms of demographics. Compared with Cluster 3, Cluster 2 comprised older patients with more comorbidities. Cluster 1 represented the group with the most severe clinical presentation, as inferred by the highest rate of hypoxemia and the highest radiological burden. Intensive care unit (ICU) admission and mechanical ventilation risks were the highest in Cluster 1. Using only two to four decision rules, the classification and regression tree (CART) phenotype assignment model achieved an AUC of 84% (81.5-86.5%, 95 CI) on the validation set. We conducted a multidimensional phenotypic analysis of adult inpatients with COVID-19 and identified three distinct phenotypes associated with different clinical outcomes. We also demonstrated the clinical usability of this approach, as phenotypes can be accurately assigned using a simple decision tree. Further research is still needed to properly incorporate these phenotypes in the management of patients with COVID-19.

Abstract 4623: Single-cell spatial proteomic analysis of the tumor microenvironment in treatment-naive NSCLC samples with immunotherapy treatment and response data

Abstract The tumor microenvironment (TME) represents a complex network comprising a variety of cell types including immune, stromal, and extracellular elements in addition to malignant tumor cells. Growing evidence suggests that the spatial relationships of these components influence tumor progression and response/resistance to immunotherapeutic agents. However, the precise mechanisms driving this relationship to clinical outcomes remains poorly understood. This is in part because generating comprehensive spatial datasets has only recently been made possible due to innovations in instrumentation, laboratory techniques, and data storage and processing. Thus, there has been little integration between various types of -omics datasets with spatial data. Here, we present a workflow for analyzing multi-omic data of the TME with the aim of providing mechanistic insights into signatures of response to immunotherapy. The workflow starts with data generation and proceeds through cell segmentation, quality filtering, phenotype assignment through unsupervised clustering, analysis of pairwise interactions and higher order multicellular structures (i.e., “neighborhoods”), and finally association with patient clinical and molecular data allowing for meaningful comparisons to be made. This analysis can be performed through the Enable Medicine Cloud Platform, a framework that facilitates the association of a multitude of data types to samples, rapid processing of multiplex images to deliver single cell spatial proteomic data, and graphical interfaces for each analysis step. To demonstrate the utility of this workflow, we generated spatial proteomic data for a tissue microarray (TMA) comprising a cohort of treatment naïve NSCLC samples (n=42) with second-line immune checkpoint inhibitor-treatment and clinical follow-up information. The spatial proteomics data was generated with the Akoya PhenoCycler-Fusion multiplexed tissue imaging platform, with an optimized panel of 51 oligonucleotide-conjugated antibodies covering a wide variety of immune, epithelial, stromal, and malignant cell markers developed by Enable Medicine. This proteomic data was then integrated with GeoMX DSP, bulk RNAseq, and clinical information using the Enable Platform. This framework for analysis allows for the comprehensive integration of both clinical and biological parameters to enable biomarker discovery and insight into signatures of response to immunotherapy. Acknowledgement: Pasteur Hospital, 30 Voie Romaine, 06000 Nice, France for NSCLC tumor samples. Citation Format: Marie Cumberbatch, Geoffrey Ivison, Amy Lam, Aaron Mayer, Milan Bhagat. Single-cell spatial proteomic analysis of the tumor microenvironment in treatment-naive NSCLC samples with immunotherapy treatment and response data. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4623.

Prospective Study of Immune Phenotype and Associated Outcomes in Patients with Leukemia and Acute Respiratory Distress Syndrome

Background: Patients with active malignancy are frequently admitted to intensive care units (ICUs) and experience organ failure including acute respiratory distress syndrome (ARDS). In ARDS clinical trial populations, two immune phenotypes have been identified which seem to differ based on plasma protein expression, predicted mortality, and response to several randomized therapies. It is unknown if similar patters exist among ICU patients with hematologic malignancies as these patients are often excluded from clinical trials in critical care. We therefore sought to classify patients with leukemia who develop ARDS secondary to septic insult according to validated immune phenotyping algorithms and describe their characteristics and outcomes. Methods: We conducted a prospective cohort study of sepsis subjects admitted to the ICU. Eligible patients had suspected infection and organ injury upon ICU admission; exclusion criteria included non-infectious causes of organ failure, a desire for comfort measures only, or lack of informed consent. Plasma collected on day 0 of ICU admission was profiled for proteins including IL8 and sTNFR1 by electrochemiluminescence. We limited our investigation to subjects with a diagnosis of active leukemia (acute or chronic) who developed ARDS (by Berlin criteria) within 6 days of ICU admission. Using a published parsimonious algorithm incorporating plasma IL8, sTNFR1, and bicarbonate, we assigned either a hyper- or hypo-inflammatory ARDS immune phenotype. We then constructed a multivariable Cox proportional hazards model controlling for APACHE III, history of BMT, sex, acute vs. chronic leukemia, and site of infection to assess the independent effect of phenotype assignment on mortality. Univariate comparisons between patients assigned to either phenotype were made as appropriate with T or Chi-Square tests. Results: 86 subjects diagnosed with leukemia and ARDS between 2008 and 2017 were identified. In-hospital mortality was 76%. 64 (74.4%) were assigned a hyper-inflammatory phenotype using the 3-marker algorithm. Male sex, higher APACHE III, and non-pulmonary infectious source appeared to be associated with hyper-inflammatory assignment in univariate comparison, while tumor-specific characteristics such as leukemia type and cytogenetic features were not (table 1). Median survival in the hyper-inflammatory group was 6 days vs. 20 in the hypo-inflammatory group (log-rank p = 0.04, figure 1). However in our Cox proportional hazards model ARDS immune phenotype was not significantly associated with mortality (HR 1.63, 95% CI 0.87 - 3.08, p = 0.12). A sub-analysis of patients diagnosed with AML did not reveal any association between European LeukemiaNet cytogenetic risk category and mortality, controlling for the same baseline covariates and immune phenotype. Interpretation: We identified previously described ARDS immune phenotypes among leukemic ARDS patients, though we did not observe an independent association between immune phenotype and mortality when controlling for relevant baseline covariates. At 76%, overall mortality was consistent with prior estimates in oncologic ICU patients and higher than the general population (~40%). A higher proportion of leukemia subjects were assigned to the hyper-inflammatory phenotype (74.4%) than described in non-oncologic trial populations (30-35% in 5 prior ARDS trials). Limitations include small sample size and knowledge that both malignancy and immunosuppression may affect cytokine expression such that phenotypic assignment is unreliable. Future efforts might focus on re-deriving ARDS phenotypes from a larger cohort of patients with hematologic malignancies in order to better understand how tumor interacts with the immunopathology of their critical illness. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Variability Between Clinical Pharmacogenetics Implementation Consortium (CPIC®) Guidelines and a Commercial Pharmacogenetics Laboratory in Genotype to Phenotype Interpretations For Patients Utilizing Psychotropics.

Clinical practice environments without in-house pharmacogenetic testing often rely on commercial laboratories, especially in the setting of pharmacogenetic testing intended to guide psychotropic use. There are occasionally differences in phenotype assignment and medication recommendations between commercial laboratories and the Clinical Pharmacogenetics Implementation Consortium (CPIC). This may be problematic as many institutions that implement pharmacogenetics consider CPIC to be an important source of guidelines for recommended prescribing actions based on genetics, as well as a tool towards standardizing pharmacogenetics implementation. Here, we completed a retrospective chart review of our academic health system’s (Michigan Medicine) electronic health record with the goal of comparing phenotypic assignment of CYP2D6 and CYP2C19 genotypes between the commercial pharmacogenetic lab used most at our institution, and CPIC. Ultimately, we identified 205 patients with available pharmacogenetic results from this lab. The prevalence of conflicting phenotype assignment was 28.8% for CYP2D6 and 32.2% for CYP2C19 genotypes when comparing the commercial lab to CPIC guidelines. In several cases, the phenotypic assignment differences for antidepressants led to significant differences in medication recommendations when comparing the commercial lab report and CPIC guidelines. These results may also have implications for medications outside of psychiatry with recommendations for dose adjustments based on CYP2D6 or CYP2C19 metabolizing phenotype.

Machine Learning and Pathway Analysis-Based Discovery of Metabolomic Markers Relating to Chronic Pain Phenotypes.

Recent scientific evidence suggests that chronic pain phenotypes are reflected in metabolomic changes. However, problems associated with chronic pain, such as sleep disorders or obesity, may complicate the metabolome pattern. Such a complex phenotype was investigated to identify common metabolomics markers at the interface of persistent pain, sleep, and obesity in 71 men and 122 women undergoing tertiary pain care. They were examined for patterns in d = 97 metabolomic markers that segregated patients with a relatively benign pain phenotype (low and little bothersome pain) from those with more severe clinical symptoms (high pain intensity, more bothersome pain, and co-occurring problems such as sleep disturbance). Two independent lines of data analysis were pursued. First, a data-driven supervised machine learning-based approach was used to identify the most informative metabolic markers for complex phenotype assignment. This pointed primarily at adenosine monophosphate (AMP), asparagine, deoxycytidine, glucuronic acid, and propionylcarnitine, and secondarily at cysteine and nicotinamide adenine dinucleotide (NAD) as informative for assigning patients to clinical pain phenotypes. After this, a hypothesis-driven analysis of metabolic pathways was performed, including sleep and obesity. In both the first and second line of analysis, three metabolic markers (NAD, AMP, and cysteine) were found to be relevant, including metabolic pathway analysis in obesity, associated with changes in amino acid metabolism, and sleep problems, associated with downregulated methionine metabolism. Taken together, present findings provide evidence that metabolomic changes associated with co-occurring problems may play a role in the development of severe pain. Co-occurring problems may influence each other at the metabolomic level. Because the methionine and glutathione metabolic pathways are physiologically linked, sleep problems appear to be associated with the first metabolic pathway, whereas obesity may be associated with the second.

DRPPM-EASY: A Web-Based Framework for Integrative Analysis of Multi-Omics Cancer Datasets.

Simple SummaryWith the influx of multi-omics profiling, effective integration of these data remains the bottleneck for omics-driven discovery. Thus, we developed DRPPM-EASY, an R Shiny framework for integrative multi-omics analysis of cancer datasets. Our tool enables the exploration of multi-omics data by providing a simple user interface that minimizes the need for computational experience. Furthermore, the interface can be deployed locally or on a webserver to facilitate scientific collaboration and discovery.High-throughput transcriptomic and proteomic analyses are now routinely applied to study cancer biology. However, complex omics integration remains challenging and often time-consuming. Here, we developed DRPPM-EASY, an R Shiny framework for integrative multi-omics analysis. We applied our application to analyze RNA-seq data generated from a USP7 knockdown in T-cell acute lymphoblastic leukemia (T-ALL) cell line, which identified upregulated expression of a TAL1-associated proliferative signature in T-cell acute lymphoblastic leukemia cell lines. Next, we performed proteomic profiling of the USP7 knockdown samples. Through DRPPM-EASY-Integration, we performed a concurrent analysis of the transcriptome and proteome and identified consistent disruption of the protein degradation machinery and spliceosome in samples with USP7 silencing. To further illustrate the utility of the R Shiny framework, we developed DRPPM-EASY-CCLE, a Shiny extension preloaded with the Cancer Cell Line Encyclopedia (CCLE) data. The DRPPM-EASY-CCLE app facilitates the sample querying and phenotype assignment by incorporating meta information, such as genetic mutation, metastasis status, sex, and collection site. As proof of concept, we verified the expression of TP53 associated DNA damage signature in TP53 mutated ovary cancer cells. Altogether, our open-source application provides an easy-to-use framework for omics exploration and discovery.

Revising Host Phenotypes of Sepsis Using Microbiology.

Background: There is wide heterogeneity in sepsis in causative pathogens, host response, organ dysfunction, and outcomes. Clinical and biologic phenotypes of sepsis are proposed, but the role of pathogen data on sepsis classification is unknown.Methods: We conducted a secondary analysis of the Recombinant Human Activated Protein C (rhAPC) Worldwide Evaluation in Severe Sepsis (PROWESS) Study. We used latent class analysis (LCA) to identify sepsis phenotypes using, (i) only clinical variables (“host model”) and, (ii) combining clinical with microbiology variables (e.g., site of infection, culture-derived pathogen type, and anti-microbial resistance characteristics, “host-pathogen model”). We describe clinical characteristics, serum biomarkers, and outcomes of host and host-pathogen models. We tested the treatment effects of rhAPC by phenotype using Kaplan-Meier curves.Results: Among 1,690 subjects with severe sepsis, latent class modeling derived a 4-class host model and a 4-class host-pathogen model. In the host model, alpha type (N = 327, 19%) was younger and had less shock; beta type (N=518, 31%) was older with more comorbidities; gamma type (N = 532, 32%) had more pulmonary dysfunction; delta type (N = 313, 19%) had more liver, renal and hematologic dysfunction and shock. After the addition of microbiologic variables, 772 (46%) patients changed phenotype membership, and the median probability of phenotype membership increased from 0.95 to 0.97 (P < 0.01). When microbiology data were added, the contribution of individual variables to phenotypes showed greater change for beta and gamma types. In beta type, the proportion of abdominal infections (from 20 to 40%) increased, while gamma type patients had an increased rate of lung infections (from 50 to 78%) with worsening pulmonary function. Markers of coagulation such as d-dimer and plasminogen activator inhibitor (PAI)-1 were greater in the beta type and lower in the gamma type. The 28 day mortality was significantly different for individual phenotypes in host and host-pathogen models (both P < 0.01). The treatment effect of rhAPC obviously changed in gamma type when microbiology data were added (P-values of log rank test changed from 0.047 to 0.780).Conclusions: Sepsis host phenotype assignment was significantly modified when microbiology data were added to clinical variables, increasing cluster cohesiveness and homogeneity.

Identification and validation of clinical phenotypes with prognostic implications in patients admitted to hospital with COVID-19: a multicentre cohort study

BackgroundThe clinical presentation of COVID-19 in patients admitted to hospital is heterogeneous. We aimed to determine whether clinical phenotypes of patients with COVID-19 can be derived from clinical data, to assess the reproducibility of these phenotypes and correlation with prognosis, and to derive and validate a simplified probabilistic model for phenotype assignment. Phenotype identification was not primarily intended as a predictive tool for mortality.MethodsIn this study, we used data from two cohorts: the COVID-19@Spain cohort, a retrospective cohort including 4035 consecutive adult patients admitted to 127 hospitals in Spain with COVID-19 between Feb 2 and March 17, 2020, and the COVID-19@HULP cohort, including 2226 consecutive adult patients admitted to a teaching hospital in Madrid between Feb 25 and April 19, 2020. The COVID-19@Spain cohort was divided into a derivation cohort, comprising 2667 randomly selected patients, and an internal validation cohort, comprising the remaining 1368 patients. The COVID-19@HULP cohort was used as an external validation cohort. A probabilistic model for phenotype assignment was derived in the derivation cohort using multinomial logistic regression and validated in the internal validation cohort. The model was also applied to the external validation cohort. 30-day mortality and other prognostic variables were assessed in the derived phenotypes and in the phenotypes assigned by the probabilistic model.FindingsThree distinct phenotypes were derived in the derivation cohort (n=2667)—phenotype A (516 [19%] patients), phenotype B (1955 [73%]) and phenotype C (196 [7%])—and reproduced in the internal validation cohort (n=1368)—phenotype A (233 [17%] patients), phenotype B (1019 [74%]), and phenotype C (116 [8%]). Patients with phenotype A were younger, were less frequently male, had mild viral symptoms, and had normal inflammatory parameters. Patients with phenotype B included more patients with obesity, lymphocytopenia, and moderately elevated inflammatory parameters. Patients with phenotype C included older patients with more comorbidities and even higher inflammatory parameters than phenotype B. We developed a simplified probabilistic model (validated in the internal validation cohort) for phenotype assignment, including 16 variables. In the derivation cohort, 30-day mortality rates were 2·5% (95% CI 1·4–4·3) for patients with phenotype A, 30·5% (28·5–32·6) for patients with phenotype B, and 60·7% (53·7–67·2) for patients with phenotype C (log-rank test p<0·0001). The predicted phenotypes in the internal validation cohort and external validation cohort showed similar mortality rates to the assigned phenotypes (internal validation cohort: 5·3% [95% CI 3·4–8·1] for phenotype A, 31·3% [28·5–34·2] for phenotype B, and 59·5% [48·8–69·3] for phenotype C; external validation cohort: 3·7% [2·0–6·4] for phenotype A, 23·7% [21·8–25·7] for phenotype B, and 51·4% [41·9–60·7] for phenotype C).InterpretationPatients admitted to hospital with COVID-19 can be classified into three phenotypes that correlate with mortality. We developed and validated a simplified tool for the probabilistic assignment of patients into phenotypes. These results might help to better classify patients for clinical management, but the pathophysiological mechanisms of the phenotypes must be investigated.FundingInstituto de Salud Carlos III, Spanish Ministry of Science and Innovation, and Fundación SEIMC/GeSIDA.

Machine-Learned Association of Next-Generation Sequencing-Derived Variants in Thermosensitive Ion Channels Genes with Human Thermal Pain Sensitivity Phenotypes.

Genetic association studies have shown their usefulness in assessing the role of ion channels in human thermal pain perception. We used machine learning to construct a complex phenotype from pain thresholds to thermal stimuli and associate it with the genetic information derived from the next-generation sequencing (NGS) of 15 ion channel genes which are involved in thermal perception, including ASIC1, ASIC2, ASIC3, ASIC4, TRPA1, TRPC1, TRPM2, TRPM3, TRPM4, TRPM5, TRPM8, TRPV1, TRPV2, TRPV3, and TRPV4. Phenotypic information was complete in 82 subjects and NGS genotypes were available in 67 subjects. A network of artificial neurons, implemented as emergent self-organizing maps, discovered two clusters characterized by high or low pain thresholds for heat and cold pain. A total of 1071 variants were discovered in the 15 ion channel genes. After feature selection, 80 genetic variants were retained for an association analysis based on machine learning. The measured performance of machine learning-mediated phenotype assignment based on this genetic information resulted in an area under the receiver operating characteristic curve of 77.2%, justifying a phenotype classification based on the genetic information. A further item categorization finally resulted in 38 genetic variants that contributed most to the phenotype assignment. Most of them (10) belonged to the TRPV3 gene, followed by TRPM3 (6). Therefore, the analysis successfully identified the particular importance of TRPV3 and TRPM3 for an average pain phenotype defined by the sensitivity to moderate thermal stimuli.

Multi-site investigation of strategies for the clinical implementation of CYP2D6 genotyping to guide drug prescribing

PurposeA number of institutions have clinically implemented CYP2D6 genotyping to guide drug prescribing. We compared implementation strategies of early adopters of CYP2D6 testing, barriers faced by both early adopters and institutions in the process of implementing CYP2D6 testing, and approaches taken to overcome these barriers. MethodsWe surveyed eight early adopters of CYP2D6 genotyping and eight institutions in the process oF.A.doption. Data were collected on testing approaches, return of results procedures, applications of genotype results, challenges faced, and lessons learned. ResultsAmong early adopters, CYP2D6 testing was most commonly ordered to assist with opioid and antidepressant prescribing. Key differences among programs included test ordering and genotyping approaches, result reporting, and clinical decision support. However, all sites tested for copy-number variation and nine common variants, and reported results in the medical record. Most sites provided automatic consultation and had designated personnel to assist with genotype-informed therapy recommendations. Primary challenges were related to stakeholder support, CYP2D6 gene complexity, phenotype assignment, and sustainability. ConclusionThere are specific challenges unique to CYP2D6 testing given the complexity of the gene and its relevance to multiple medications. Consensus lessons learned may guide those interested in pursuing similar clinical pharmacogenetic programs.

Distinguishing Wheezing Phenotypes from Infancy to Adolescence. A Pooled Analysis of Five Birth Cohorts.

Rationale: Pooling data from multiple cohorts and extending the time frame across childhood should minimize study-specific effects, enabling better characterization of childhood wheezing.Objectives: To analyze wheezing patterns from early childhood to adolescence using combined data from five birth cohorts.Methods: We used latent class analysis to derive wheeze phenotypes among 7,719 participants from five birth cohorts with complete report of wheeze at five time periods. We tested the associations of derived phenotypes with late asthma outcomes and lung function, and investigated the uncertainty in phenotype assignment.Results: We identified five phenotypes: never/infrequent wheeze (52.1%), early onset preschool remitting (23.9%), early onset midchildhood remitting (9%), persistent (7.9%), and late-onset wheeze (7.1%). Compared with the never/infrequent wheeze, all phenotypes had higher odds of asthma and lower forced expiratory volume in 1 second and forced expiratory volume in 1 second/forced vital capacity in adolescence. The association with asthma was strongest for persistent wheeze (adjusted odds ratio, 56.54; 95% confidence interval, 43.75–73.06). We observed considerable within-class heterogeneity at the individual level, with 913 (12%) children having low membership probability (<0.60) of any phenotype. Class membership certainty was highest in persistent and never/infrequent, and lowest in late-onset wheeze (with 51% of participants having membership probabilities <0.80). Individual wheezing patterns were particularly heterogeneous in late-onset wheeze, whereas many children assigned to early onset preschool remitting class reported wheezing at later time points.Conclusions: All wheeze phenotypes had significantly diminished lung function in school-age children, suggesting that the notion that early life episodic wheeze has a benign prognosis may not be true for a proportion of transient wheezers. We observed considerable within-phenotype heterogeneity in individual wheezing patterns.

Genotype/phenotype correlations in 538 congenital adrenal hyperplasia patients from Germany and Austria: discordances in milder genotypes and in screened versus prescreening patients.

Congenital adrenal hyperplasia (CAH) due to CYP21A2 gene mutations is associated with a variety of clinical phenotypes (salt wasting, SW; simple virilizing, SV; nonclassical, NC) depending on residual 21-hydroxylase activity. Phenotypes and genotypes correlate well in 80–90% of cases. We set out to test the predictive value of CAH phenotype assignment based on genotype classification in a large multicenter cohort. A retrospective evaluation of genetic data from 538 CAH patients (195 screened) collected from 28 tertiary centers as part of a German quality control program was performed. Genotypes were classified according to residual 21-hydroxylase activity (null, A, B, C) and assigned clinical phenotypes correlated with predicted phenotypes, including analysis of Prader stages. Ultimately, concordance of genotypes with clinical phenotypes was compared in patients diagnosed before or after the introduction of nationwide CAH-newborn screening. Severe genotypes (null and A) correlated well with the expected phenotype (SW in 97 and 91%, respectively), whereas less severe genotypes (B and C) correlated poorly (SV in 45% and NC in 57%, respectively). This was underlined by a high degree of virilization in girls with C genotypes (Prader stage >1 in 28%). SW was diagnosed in 90% of screening-positive babies with classical CAH compared with 74% of prescreening patients. In our CAH series, assigned phenotypes were more severe than expected in milder genotypes and in screened vs prescreening patients. Diagnostic discrimination between phenotypes based on genotypes may prove overcome due to the overlap in their clinical presentations.