Random Forest Classification Analysis Research Articles

Fecal microbiota and volatile metabolome pattern alterations precede late-onset meningitis in preterm neonates.

The fecal microbiota and metabolome are hypothesized to be altered before late-onset neonatal meningitis (LOM), in analogy to late-onset sepsis (LOS). The present study aimed to identify fecal microbiota composition and volatile metabolomics preceding LOM. Cases and gestational age-matched controls were selected from a prospective, longitudinal preterm cohort study (born <30 weeks' gestation) at nine neonatal intensive care units. The microbial composition (16S rRNA sequencing) and volatile metabolome (gas chromatography-ion mobility spectrometry (GC-IMS) and GC-time-of-flight-mass spectrometry (GC-TOF-MS)), were analyzed in fecal samples 1-10 days pre-LOM. Of 1397 included infants, 21 were diagnosed with LOM (1.5%), and 19 with concomitant LOS (90%). Random Forest classification and MaAsLin2 analysis found similar microbiota features contribute to the discrimination of fecal pre-LOM samples versus controls. A Random Forest model based on six microbiota features accurately predicts LOM 1-3 days before diagnosis with an area under the curve (AUC) of 0.88 (n=147). Pattern recognition analysis by GC-IMS revealed an AUC of 0.70-0.76 (P<0.05) in the three days pre-LOM (n=92). No single discriminative metabolites were identified by GC-TOF-MS (n=66). Infants with LOM could be accurately discriminated from controls based on preclinical microbiota composition, while alterations in the volatile metabolome were moderately associated with preclinical LOM.

377 CDKN2A Alterations Are Associated With Increased Postoperative Seizure Risk After Resection of Brain Metastases

INTRODUCTION: Seizures are common and significantly disabling for patients with brain metastases (BMs). While surgical resection can provide seizure control, a subset of patients with BMs may continue to suffer seizures postoperatively. Molecular BM characteristics may influence which patients are at-risk for postoperative seizures. METHODS: All patients underwent BM resection at one institution with available clinical and sequencing data on &gt; 500 oncogenes. Clinical seizures were documented preoperatively and postoperatively. Random forest machine learning classification was used to determine candidate genomic alterations associated with postoperative seizures, and clinical and top genomic variables were correlated with postoperative seizures using Cox proportional hazards models. RESULTS: There were 112 patients with BMs who underwent 114 surgeries and had at least one month of postoperative follow-up. Seizures occurred preoperatively in 26 (22.8%) patients and postoperatively in 25 (21.9%). Engel classification (EC) at six months achieved for those with preoperative seizures was EC I in 13 (50%); II in 6 (23.1%); III in 5 (19.2%), and IV in 2 (7.7%). In those with postoperative seizures, only 8 (32.0%) had seizures preoperatively. Preoperative seizures were not a significant predictor of postoperative seizures (hazards ratio (HR) = 1.84; 95% confidence interval (CI) = 0.79 – 4.37; p = 0.156). On random forest classification and multivariate Cox analysis controlling for factors including recurrence, extent of resection, and number of BMs, CDKN2A alterations were associated with postoperative seizures (HR = 3.22; 95% CI = 1.27–8.16; p = 0.014). Melanoma BMs were associated with higher risk of postoperative seizures compared with all other malignancies (HR = 5.23; 95% CI = 1.37–19.98; p = 0.016). CONCLUSIONS: CDKN2A alterations and melanoma primary malignancy are associated with increased postoperative seizure risk following resection of BMs. These results may help guide postoperative seizure prophylaxis in patients undergoing resection of BMs.

Development of a 3–step sequential extraction method to investigate the fraction and affecting factors of 21 antibiotics in soils

Antibiotic exist in various states after entering agricultural soil through the application of manure, including the aqueous state (I), which can be directly absorbed by plants, and the auxiliary organic extraction state (III), which is closely associated with the pseudo-permanence of antibiotics. However, effective analytical methods for extracting and affecting factors on fractions of different antibiotic states remain unclear. In this study, KCl, acetonitrile/Na2EDTA-McIlvaine buffer, and acetonitrile/water were successively used to extract states I, II, and III of 21 antibiotics in soil, and the recovery efficiency met the quantitative requirements. Random forest classification and variance partitioning analysis revealed that dissolved organic matter, pH, and organic matter were important factors affecting the recovery efficiency of antibiotic in states I, II, and III, respectively. Additionally, 65-day spiked soil experiments combined with Mantel test analysis suggested that pH, organic acids, heavy metals, and noncrystalline minerals differentially affected antibiotic type and state. Importantly, a structural equation model indicated that organic acids play a crucial role in the fraction of antibiotic states. Overall, this study reveals the factors influencing the fraction of different antibiotic states in soil, which is helpful for accurately assessing their ecological risk.

Whistle repertoire and structure reflect ecotype distinction of pantropical spotted dolphins in the Eastern Tropical Pacific

The pantropical spotted dolphin in the Eastern Tropical Pacific (ETP) is found in two genetically and phenotypically diverged ecotypes, coastal and offshore. These habitats have distinct acoustic characteristics, which can lead to the evolution of distinct acoustic communication. Whistles are sounds widely used by dolphins to mediate species and individual recognition and social interactions. Here, we study the whistle acoustic structure and repertoire diversity of offshore and coastal pantropical spotted dolphins. Our results show that there is significantly more within- and across-group variation in whistle fundamental frequency between ecotypes than between offshore groups and between coastal groups. A Random Forest classification analysis performed with an accuracy of 83.99% and identified duration, peak and minimum frequency as the most informative variables for distinguishing between ecotypes. Overall, coastal spotted dolphins produced significantly shorter whistles that were significantly lower in frequency (peak, minimum and maximum, and start and end) than offshore dolphins. Ecotypes produced whistle repertoires that were similar in diversity, but different in contour composition, with the coastal ecotype producing more upsweep whistles than offshore dolphins. The results of this study suggest that acoustic adaptations to coastal and offshore environments could be important contributors to intraspecific variation of dolphin whistle repertoires.

Coordinators in the return-to-work process: Mapping their work models.

In recent decades, many countries have implemented return-to-work coordinators to combat high rates of sickness absence and insufficient collaboration in the return-to-work process. The coordinators should improve communication and collaboration between stakeholders in the return-to-work process for people on sickness absence. How they perform their daily work remains unexplored, and we know little about to what extent they collaborate and perform other work tasks to support people on sickness absence. This study examines which work models return-to-work coordinators use in primary healthcare, psychiatry and orthopaedics in Sweden. A questionnaire was sent to all 82 coordinators in one region (89% response rate) with questions about the selection of patients, individual patient support, healthcare collaboration, and external collaboration. Random forest classification analysis was used to identify the models. Three work models were identified. In model A, coordinators were more likely to select certain groups of patients, spend more time in telephone than in face-to-face meetings, and collaborate fairly much. In Model B there was less patient selection and much collaboration and face-to-face meetings. Model C involved little patient selection, much telephone contact and very little collaboration. Model A was more common in primary healthcare, model C in orthopaedics, while model B was distributed equally between primary healthcare and psychiatry. The work models correspond differently to the coordinator's assignments of supporting patients and collaborating with healthcare and other stakeholders. The differences lie in how much they actively select patients, how much they collaborate, and with whom. Their different distribution across clinical contexts indicates that organisational demands influence how work models evolve in practice.

Association of CDKN2A alterations with increased postoperative seizure risk after resection of brain metastases.

Seizures are common and significantly disabling for patients with brain metastases (BMs). Although resection can provide seizure control, a subset of patients with BMs may continue to suffer seizures postoperatively. Genomic BM characteristics may influence which patients are at risk for postoperative seizures. This work explores correlations between genomic alterations and risk of postoperative seizures following BM resection. All patients underwent BM resection at a single institution, with available clinical and sequencing data on more than 500 oncogenes. Clinical seizures were documented pre- and postoperatively. A random forest machine learning classification was used to determine candidate genomic alterations associated with postoperative seizures, and clinical and top genomic variables were correlated with postoperative seizures by using Cox proportional hazards models. There were 112 patients with BMs who underwent 114 surgeries and had at least 1 month of postoperative follow-up. Seizures occurred preoperatively in 26 (22.8%) patients and postoperatively in 25 (21.9%). The Engel classification achieved at 6 months for those with preoperative seizures was class I in 13 (50%); class II in 6 (23.1%); class III in 5 (19.2%), and class IV in 2 (7.7%). In those with postoperative seizures, only 8 (32.0%) had seizures preoperatively, and preoperative seizures were not a significant predictor of postoperative seizures (HR 1.84; 95% CI 0.79-4.37; p = 0.156). On random forest classification and multivariate Cox analysis controlling for factors including recurrence, extent of resection, and number of BMs, CDKN2A alterations were associated with postoperative seizures (HR 3.22; 95% CI 1.27-8.16; p = 0.014). Melanoma BMs were associated with higher risk of postoperative seizures compared with all other primary malignancies (HR 5.23; 95% CI 1.37-19.98; p = 0.016). Of 39 BMs with CDKN2A alteration, 35.9% (14/39) had postoperative seizures, compared to 14.7% (11/75) without CDKN2A alteration. The overall rate of postoperative seizures in melanoma BMs was 42.9% (15/35), compared with 12.7% (10/79) for all other primary malignancies. CDKN2A alterations and melanoma primary malignancy are associated with increased postoperative seizure risk following resection of BMs. These results may help guide postoperative seizure prophylaxis in patients undergoing resection of BMs.

Assessing the accuracy of California county level COVID-19 hospitalization forecasts to inform public policy decision making

BackgroundThe COVID-19 pandemic has highlighted the role of infectious disease forecasting in informing public policy. However, significant barriers remain for effectively linking infectious disease forecasts to public health decision making, including a lack of model validation. Forecasting model performance and accuracy should be evaluated retrospectively to understand under which conditions models were reliable and could be improved in the future.MethodsUsing archived forecasts from the California Department of Public Health’s California COVID Assessment Tool (https://calcat.covid19.ca.gov/cacovidmodels/), we compared how well different forecasting models predicted COVID-19 hospitalization census across California counties and regions during periods of Alpha, Delta, and Omicron variant predominance.ResultsBased on mean absolute error estimates, forecasting models had variable performance across counties and through time. When accounting for model availability across counties and dates, some individual models performed consistently better than the ensemble model, but model rankings still differed across counties. Local transmission trends, variant prevalence, and county population size were informative predictors for determining which model performed best for a given county based on a random forest classification analysis. Overall, the ensemble model performed worse in less populous counties, in part because of fewer model contributors in these locations.ConclusionsEnsemble model predictions could be improved by incorporating geographic heterogeneity in model coverage and performance. Consistency in model reporting and improved model validation can strengthen the role of infectious disease forecasting in real-time public health decision making.

A high-salt diet induces synaptic loss and memory impairment via gut microbiota and butyrate in mice.

High-salt diet (HSD)-fed mice display cognitive impairment and lower synaptic proteins via changed gut microbiota composition and short-chain fatty acids production. Gut microbiota from HSD-fed mice impairs memory and synapse in normal salt diet-fed mice. Butyrate treatment partially reverses memory impairment in HSD-fed mice. Above all, this study indicates the important role of the gut microbiome and butyrate production in synaptic loss and memory impairment.

Effects of Chinese Milk Vetch Returning on Soil Properties, Microbial Community, and Rice Yield in Paddy Soil

The application of green manure is considered a sustainable management approach to improve soil quality and crop yield in agricultural production. However, few studies have focused on the effects on soil properties and the microbial community when varied green manure rates are applied in rice cropping systems. We examined the soil and microbial properties in paddy soil and evaluated rice productivity and sustainable yield index in response to Chinese milk vetch (MV) application at various rates (0, 22.5, 30, 45, and 60 Mg ha−1, labeled MV0, MV22.5, MV30, MV45, and MV60, respectively) in a seven-year (2014–2020) field experiment. For all treatments, MV application significantly improved rice yield by 22.5–29.7% and sustainable yield index by 10.3–13.0% compared with no MV application, and the highest average yield across the 7 years of the study was found in MV60, followed by MV45, MV30, and MV22.5. Compared with MV0, MV45 and MV60 treatments significantly reduced soil bulk density (BD), increased soil organic carbon (SOC), total nitrogen (TN), total potassium (TK), and alkali solution nitrogen (AN), particularly in MV45. Moreover, compared with MV0, a marked decrease in the alpha diversity (Shannon index) of the bacterial community and a significant increase in the relative abundance of Firmicutes were observed under MV45. The alpha diversity of the fungal community did not vary across treatments, but MV45 significantly reduced the relative abundance of Ascomycota and Chytridiomycota compared with MV0. In addition, redundancy analysis indicated that the fungal community structure was significantly related to soil BD, and random forest classification analysis demonstrated that SOC and AN were the most important factors for rice yield in this study. Overall, these results indicated that MV application at 45 Mg ha−1 should be recommended to simultaneously achieve the improvement of rice yield and soil quality in the South China.

Applications of Multivariate Statistical and Data Mining Analyses to the Search for Biomarkers of Sensorineural Hearing Loss, Tinnitus, and Vestibular Dysfunction.

Disorders of sensory systems, as with most disorders of the nervous system, usually involve the interaction of multiple variables to cause some change, and yet often basic sensory neuroscience data are analyzed using univariate statistical analyses only. The exclusive use of univariate statistical procedures, analyzing one variable at a time, may limit the potential of studies to determine how interactions between variables may, as a network, determine a particular result. The use of multivariate statistical and data mining methods provides the opportunity to analyse many variables together, in order to appreciate how they may function as a system of interacting variables, and how this system or network may change as a result of sensory disorders such as sensorineural hearing loss, tinnitus or different types of vestibular dysfunction. Here we provide an overview of the potential applications of multivariate statistical and data mining techniques, such as principal component and factor analysis, cluster analysis, multiple linear regression, random forest regression, linear discriminant analysis, support vector machines, random forest classification, Bayesian classification, and orthogonal partial least squares discriminant analysis, to the study of auditory and vestibular dysfunction, with an emphasis on classification analytic methods that may be used in the search for biomarkers of disease.

Spectral contrast reduction in Australian English /l/-final rimes.

Vowel contrasts may be reduced or neutralized before coda laterals in English [Bernard (1985). The Cultivated Australian: Festschrift in Honour of Arthur Delbridge, pp. 319-332; Labov, Ash, and Boberg (2008). The Atlas of North American English, Phonetics and Sound Change (Gruyter Mouton, Berlin); Palethorpe and Cox (2003). International Seminar on Speech Production (Macquaire University, Sydney, Australia)], but the acoustic characteristics of vowel-lateral interaction in Australian English (AusE) rimes have not been systematically examined. Spectral and temporal properties of 16 pre-lateral and 16 pre-obstruent vowels produced by 29 speakers of AusE were compared. Acoustic vowel similarity in both environments was captured using random forest classification and hierarchical cluster analysis of the first three DCT coefficients of F1, F2, and F3, and duration values. Vowels preceding /l/ codas showed overall increased confusability compared to vowels preceding /d/ codas. In particular, reduced spectral contrast was found for the rime pairs /iːl-ɪl/ (feel-fill), /ʉːl-ʊl/ (fool-full), /əʉl-ɔl/ (dole-doll), and /æɔl-æl/ (howl-Hal). Potential articulatory explanations and implications for sound change are discussed.

Who's onboard? A predictive analysis of cooperative formation in commercial fisheries

Fishing cooperatives around the world have increasingly taken on co-management of commercial fisheries in recent decades, with generally positive results in meeting management targets and increasing economic value. To better understand which commercial fisheries or fleets are likely to form cooperative associations in the future, we utilized a database of management and fleet-level attributes for 106 fisheries—mainly industrial fisheries from the United States, New Zealand, Canada, and the United Kingdom—to develop a predictive model. We considered two alternative definitions of cooperatives: a legal, operational definition that classified 51 of the fisheries as cooperatives, and a more stringent proactive definition that classified 35 of the fisheries as cooperatives. Random forest classification analyses showed that cooperatives of both types were most likely to form in fisheries with greater boat cost, greater level of participation in industry associations, and in fisheries with bycatch limits; strong regional effects were also observed. Cross-validation prediction accuracy levels were high: using 10 predictor variables, 86% and 91% of fisheries were correctly classified under the operational and proactive cooperative definitions, respectively. These predictions suggest which fisheries may be next to create cooperative fishing associations or engage in more proactive arrangements within cooperatives. These results point to which regulatory reforms, such as flexible bycatch restrictions, could lead to more cooperative behavior in fisheries.

Association of the Psoriatic Microenvironment With Treatment Response

The ability to predict the efficacy of systemic psoriasis therapy based on immune profiles in skin biopsies could reduce the use of inappropriate treatment and its associated costs and adverse events. It could considerably decrease drug development trial costs as well. To develop a bioinformatic gene signature score derived from skin mRNA to predict psoriasis treatment outcomes for a variety of therapies. In this decision analytical model using 1145 skin samples from different cohorts of 12 retrospective psoriasis studies, samples were analyzed using the CIBERSORT algorithm to define the immune landscape of psoriasis lesions and controls. Random forest classification and principal component analysis algorithms were used to estimate psoriatic microenvironment (PME) signature genes and construct a PME score. Overall, 85 and 421 psoriasis lesions from 1 and 4 independent cohorts were used as discovery and validation studies, respectively. Among them, 157, 71, 89, and 90 psoriasis lesions were treated with etanercept, tofacitinib, adalimumab, and methotrexate, respectively. Number of weeks after treatment initiation when responders and nonresponders could be predicted. Overall, 22 immune cell subtypes formed infiltration patterns that differentiated psoriasis lesions from healthy skin. In psoriasis lesions, the expression of 33 PME signature genes defined 2 immune phenotypes and in aggregate could be simplified to a numerical PME score. A high PME score, characterized by keratinocyte differentiation, correlated with a better treatment response (Psoriasis Area and Severity Index [PASI] reduction, 75.8%; 95% CI, 69.4% to 82.2%; P = .03), whereas a low PME score exhibited an immune activation signature and was associated with a worse response (PASI reduction, 53.5%; 95% CI, 45.3% to 61.7%; P = .03). The PME score at week 4 after treatment initiation correlated with future responder vs nonresponder to treatment status 8 to 12 weeks earlier than PASI reduction for etanercept, methotrexate plus adalimumab, and tofacitinib. The PME score is a biometric score that may predict clinical efficacy of systemic psoriasis therapy in advance of clinical responses. As an application of personalized medicine, it may reduce the exposure of patients with psoriasis to ineffective and expensive therapies.

MenACWY vaccination campaign for adolescents in the Netherlands: Uptake and its determinants

BackgroundBetween 2015 and 2018 the incidence of meningococcal disease serogroup W increased tenfold in the Netherlands with the highest case fatality among adolescents and young adults. In 2018–2019 a national mass vaccination campaign, with recall after non-attendance, was implemented targeting 14–18-years old adolescents. This study estimated the MenACWY-vaccine uptake and investigated its determinants. MethodsThe uptake before the start of the vaccination campaign was estimated from the number of vaccines administered by Municipal Health Services and dispensed by public pharmacies, and within the campaign from the national vaccination register. Possible determinants of uptake after the first invitation and recall were investigated among the first group invited for vaccination (born in May-December 2004) using random forest classification analysis. ResultsThe uptake was 86% (of which 1.9% before the campaign) among all eligible adolescents and 88% among the first group invited, including 5% uptake achieved after the recall. The most important predictor of vaccination after the first invitation was parents’ country of birth (lower uptake when parents were born abroad, range: 52%-Morocco to 88%-Netherlands). The most important predictors among those recalled were, respectively, distance to vaccination location (lower uptake with larger distance, range: 4–6%), percentage of votes for the conservative Christian (reformed) party in the municipality (lower uptake with higher percentage, range: 4–5%) and parents’ country of birth (higher uptake when parents were born abroad, range: 4%-Netherlands to 11%-Syria). ConclusionsThe MenACWY vaccination campaign for adolescents achieved a high uptake, with little vaccine use before the campaign. Parents’ country of birth, votes for the conservative reformed political party and distance to the vaccination location were the most important predictors of vaccination. The recall strategy enhanced the uptake and was valuable to diminish immunization disparities. Future vaccination campaigns should put more effort into reaching adolescents with immigrant parents.

Long-Term Land Use/Land Cover Change Assessment of the Kilombero Catchment in Tanzania Using Random Forest Classification and Robust Change Vector Analysis

Information about land use/land cover (LULC) and their changes is useful for different stakeholders to assess future pathways of sustainable land use for food production as well as for nature conservation. In this study, we assess LULC changes in the Kilombero catchment in Tanzania, an important area of recent development in East Africa. LULC change is assessed in two ways: first, post-classification comparison (PCC) which allows us to directly assess changes from one LULC class to another, and second, spectral change detection. We perform LULC classification by applying random forests (RF) on sets of multitemporal metrics that account for seasonal within-class dynamics. For the spectral change detection, we make use of the robust change vector analysis (RCVA) and determine those changes that do not necessarily lead to another class. The combination of the two approaches enables us to distinguish areas that show (a) only PCC changes, (b) only spectral changes that do not affect the classification of a pixel, (c) both types of change, or (d) no changes at all. Our results reveal that only one-quarter of the catchment has not experienced any change. One-third shows both, spectral changes and LULC conversion. Changes detected with both methods predominantly occur in two major regions, one in the West of the catchment, one in the Kilombero floodplain. Both regions are important areas of food production and economic development in Tanzania. The Kilombero floodplain is a Ramsar protected area, half of which was converted to agricultural land in the past decades. Therefore, LULC monitoring is required to support sustainable land management. Relatively poor classification performances revealed several challenges during the classification process. The combined approach of PCC and RCVA allows us to detect spatial patterns of LULC change at distinct dimensions and intensities. With the assessment of additional classifier output, namely class-specific per-pixel classification probabilities and derived parameters, we account for classification uncertainty across space. We overlay the LULC change results and the spatial assessment of classification reliability to provide a thorough picture of the LULC changes taking place in the Kilombero catchment.

Infarction with associated pseudosarcomatous changes mimics anaplasia in otherwise grade I meningiomas.

We describe a morphologically distinct pattern of tumor infarction and associated sarcoma-like changes, mimicking focal anaplasia, in otherwise WHO grade I meningiomas. The described cases (n = 9) all demonstrated a discrete spindle-cell (pseudosarcomatous) component with brisk mitotic activity (12-14 mitoses/10 HPF), elevated Ki-67 (mean 75.5 ± 25.0%, quantified), absence of PR, SSTR2A, or EMA expression, and potential SMA expression (50%). Despite these high-grade features, all nine patients remained free of progression or recurrence post resection (follow-up mean: 49.8 months). In contrast, among a comparison (control) cohort of consecutive WHO grade II and III meningiomas (n = 16), as expected, progression rate was high (68.8%, P = 0.002, Fisher's exact, average time to progression = 25 months, follow-up mean: 39.8 months). While necrosis was a frequent feature among atypical/anaplastic meningiomas (12/16, 75%), and elevated mitoses and proliferative index were present consistent with histologic grade, a well-defined zonal pattern with pseudosarcomatous component was not present among these tumors. DNA methylation-based analysis readily distinguished meningiomas by copy number profiles and DNA-based methylation meningioma random forest classification analysis (meningioma v2.4 classifier developed at University of Heidelberg); all pseudosarcomatous cases analyzed (4/9) matched with high level calibrated classifier score to "MC benign-1", with isolated loss of chromosome 22q identified as the sole copy number alteration. In contrast, multiple chromosomal losses were detected among the comparison cohort and classifier results demonstrated good concordance with histologic grade. Our findings suggest that pseudosarcomatous alterations represent reactive changes to central meningioma infarction, rather than focal anaplasia, and further support the use of DNA methylation-based analysis as a useful adjunct for predicting meningioma behavior. These indolent tumors should be distinguished from their atypical and anaplastic counterparts.

The subway microbiome: seasonal dynamics and direct comparison of air and surface bacterial communities

BackgroundMass transit environments, such as subways, are uniquely important for transmission of microbes among humans and built environments, and for their ability to spread pathogens and impact large numbers of people. In order to gain a deeper understanding of microbiome dynamics in subways, we must identify variables that affect microbial composition and those microorganisms that are unique to specific habitats.MethodsWe performed high-throughput 16S rRNA gene sequencing of air and surface samples from 16 subway stations in Oslo, Norway, across all four seasons. Distinguishing features across seasons and between air and surface were identified using random forest classification analyses, followed by in-depth diversity analyses.ResultsThere were significant differences between the air and surface bacterial communities, and across seasons. Highly abundant groups were generally ubiquitous; however, a large number of taxa with low prevalence and abundance were exclusively present in only one sample matrix or one season. Among the highly abundant families and genera, we found that some were uniquely so in air samples. In surface samples, all highly abundant groups were also well represented in air samples. This is congruent with a pattern observed for the entire dataset, namely that air samples had significantly higher within-sample diversity. We also observed a seasonal pattern: diversity was higher during spring and summer. Temperature had a strong effect on diversity in air but not on surface diversity. Among-sample diversity was also significantly associated with air/surface, season, and temperature.ConclusionsThe results presented here provide the first direct comparison of air and surface bacterial microbiomes, and the first assessment of seasonal variation in subways using culture-independent methods. While there were strong similarities between air and surface and across seasons, we found both diversity and the abundances of certain taxa to differ. This constitutes a significant step towards understanding the composition and dynamics of bacterial communities in subways, a highly important environment in our increasingly urbanized and interconnect world.AZfigM7Ccn69mJ5i-QZDFtVideo abstract.

MENTAL HEALTH AND HUNGER: RISK FACTORS FOR HOSPITALIZATION AMONG HOME-DELIVERED MEAL PARTICIPANTS

Home-delivered meals (HDMs) provided through the Older Americans Act (OAA) are intended to reduce hunger, promote socialization, and maximize wellness. However, HDM recipients are at increased likelihood of being hospitalized due to their complex health needs and risk for social isolation and depression. Drawing data from the OAA-HDM National Survey, we evaluated the predictors of hospitalization among HDM recipients in 2017. From our sample (n = 578), we conducted random forest classification analyses to identify the most important risk factors related to HDM recipient hospitalization. Our random forest model yielded an accuracy rate of 66.3% with risk factors most indicative of hospitalization being attributed to number of co-morbidities, depressive symptoms, and feelings of social isolation. These findings indicate that although HDMs may help alleviate hunger among older adults, innovate strategies are warranted to address the unmet mental health needs of HDM recipients.

Abstract 1795: MicroRNA sequence analysis reveals signatures of oral cancer lymph node metastasis

Abstract Regional lymph node metastasis (LN+) at the time of diagnosis or during post-treatment follow-up has devastating impact on the decimal survival of oral cancer (OC) patients. Early intervention on at-risk patient may improve survival. The purpose of the study was to identify microRNA (miRNA) signatures that contribute to LN+ using miRNA sequencing in OC surgical samples. The study comprised a Discovery Cohort (n=91) and a Validation Cohort (n=67) of fresh-frozen primary tumors from 158 patients who were surgically treated between 2005 and 2016 with known outcome of LN (LN+, n=76; LN0, n=82). Total RNA was harvested from &amp;gt;70% tumoral areas. MiRNA-seq libraries were constructed from 2µg total RNA per sample followed by adaptor ligation, reverse transcription, PCR, and sequencing on HiSeq 2000 platform. All analyses were performed using R package (v3.4.4) with input matrix of the miRNAs that expressed at a level of at least 10 reads per million (RPM) in at least 10% of samples. A Least Absolute Shrinkage and Selection Operator (LASSO) penalized regression approach was used to identify miRNAs that would best predict nodal-disease free survival (NFS). Random forest classification analysis was used to rank miRNAs for classifying samples as LN0 and LN+. Among the 158 samples, the LN+ group consisted more poorly differentiated tumors and greater depth of invasion (p=0.0001 and 0.002, respectively). To identify miRNAs that contribute to LN+, we used a penalized regression to generate a signature of NFS. The Discovery Cohort was randomly partitioned into training set (n=72; LN+, n=39, LN0, n=33) to generate a predictive model on NFS. This was then tested in the remaining 19 cases (LN+, n=12, LN0, n=7). The model composed of 2 miRNAs (miR-107 and miR-21-5p) which were significantly overexpressed in LN+. When this was tested in the Validation Cohort (n=67; LN+, 25; LN0, 42), the model was significantly associated with NFS (HR=3.4; 95% CI, 1.5-7.6; 5-year NFS, 24.5%, p=0.001). Additionally, we ranked miRNAs for classifying all samples into LN+ and LN0 by using random forest classification analysis and the two miRNAs from the penalized regression analysis remained to be the top two. The classified LN+ group had significant inferior NFS compared to classified LN0 group (HR=3.2; 95% CI, 2.0-5.1; 5-year NFS, 27.8% vs. 66.7%, p&amp;lt;0.0001). We further investigated diagnostic performance of the two miRNAs using receiver operating curve on all samples which had accuracy of 0.88, sensitivity of 0.82, specificity of 0.80, positive predictive value of 0.82, and negative predictive value of 0.80. In conclusion, our data demonstrated noticeable dysregulated miRNAs that may serve as predictive biomarkers for LN+. Note: This abstract was not presented at the meeting. Citation Format: Kelly Yp Liu, Denise Brooks, Reanne Bowlby, Steve Jones, Catherine F. Poh. MicroRNA sequence analysis reveals signatures of oral cancer lymph node metastasis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1795.

Identification and validation of genetic variants predictive of gait in standardbred horses.

Several horse breeds have been specifically selected for the ability to exhibit alternative patterns of locomotion, or gaits. A premature stop codon in the gene DMRT3 is permissive for “gaitedness” across breeds. However, this mutation is nearly fixed in both American Standardbred trotters and pacers, which perform a diagonal and lateral gait, respectively, during harness racing. This suggests that modifying alleles must influence the preferred gait at racing speeds in these populations. A genome-wide association analysis for the ability to pace was performed in 542 Standardbred horses (n = 176 pacers, n = 366 trotters) with genotype data imputed to ~74,000 single nucleotide polymorphisms (SNPs). Nineteen SNPs on nine chromosomes (ECA1, 2, 6, 9, 17, 19, 23, 25, 31) reached genome-wide significance (p < 1.44 x 10−6). Variant discovery in regions of interest was carried out via whole-genome sequencing. A set of 303 variants from 22 chromosomes with putative modifying effects on gait was genotyped in 659 Standardbreds (n = 231 pacers, n = 428 trotters) using a high-throughput assay. Random forest classification analysis resulted in an out-of-box error rate of 0.61%. A conditional inference tree algorithm containing seven SNPs predicted status as a pacer or trotter with 99.1% accuracy and subsequently performed with 99.4% accuracy in an independently sampled population of 166 Standardbreds (n = 83 pacers, n = 83 trotters). This highly accurate algorithm could be used by owners/trainers to identify Standardbred horses with the potential to race as pacers or as trotters, according to the genotype identified, prior to initiating training and would enable fine-tuning of breeding programs with designed matings. Additional work is needed to determine both the algorithm’s utility in other gaited breeds and whether any of the predictive SNPs play a physiologically functional role in the tendency to pace or tag true functional alleles.