Blind Test Set Research Articles

Rapid detection of colistin resistance protein MCR-1 by LC\u2013MS/MS

BackgroundColistin (polymyxin E) and polymixin B are important bactericidal antibiotics used in the treatment of serious infections caused by multi-drug resistant Gram-negative organisms. Transferrable plasmid-mediated colistin resistance, conferred by the product of the mcr-1 gene, has emerged as a global healthcare threat. Consequently, the rapid detection of the MCR-1 protein in clinical bacterial isolates has become increasingly important. We used a genoproteomic approach to identify unique peptides of the MCR-1 protein that could be detected rapidly by liquid chromatography tandem mass spectrometry (LC–MS/MS).MethodsMCR-1 tryptic peptides that were efficiently ionized and readily detectable were characterized in a set of mcr-1-containing isolates with triple quadrupole LC–MS. Three optimal peptides were selected for the development of a rapid multiple reaction monitoring LC–MS/MS assay for the MCR-1 protein. To investigate the feasibility of rapid detection of the MCR-1 protein in bacterial isolates using this assay, a blinded 99-sample test set was built that included three additional mcr-1-containing clinical isolates tested in triplicate (9 samples) and 90 negative control isolates.ResultsAll of the mcr-1-containing isolates in the test set were accurately identified with no false positive detections by three independent, blinded operators, yielding an overall performance of 100% sensitivity and specificity for multiple operators. Among the three peptides tested in this study, the best performing was DTFPQLAK. The isolate-to-result time for the assay as implemented is less than 90 min.ConclusionsThis work demonstrates the feasibility of rapid detection of the MCR-1 protein in bacterial isolates by LC–MS/MS.

Identifying mouse developmental essential genes using machine learning.

ABSTRACTThe genes that are required for organismal survival are annotated as ‘essential genes’. Identifying all the essential genes of an animal species can reveal critical functions that are needed during the development of the organism. To inform studies on mouse development, we developed a supervised machine learning classifier based on phenotype data from mouse knockout experiments. We used this classifier to predict the essentiality of mouse genes lacking experimental data. Validation of our predictions against a blind test set of recent mouse knockout experimental data indicated a high level of accuracy (>80%). We also validated our predictions for other mouse mutagenesis methodologies, demonstrating that the predictions are accurate for lethal phenotypes isolated in random chemical mutagenesis screens and embryonic stem cell screens. The biological functions that are enriched in essential and non-essential genes have been identified, showing that essential genes tend to encode intracellular proteins that interact with nucleic acids. The genome distribution of predicted essential and non-essential genes was analysed, demonstrating that the density of essential genes varies throughout the genome. A comparison with human essential and non-essential genes was performed, revealing conservation between human and mouse gene essentiality status. Our genome-wide predictions of mouse essential genes will be of value for the planning of mouse knockout experiments and phenotyping assays, for understanding the functional processes required during mouse development, and for the prioritisation of disease candidate genes identified in human genome and exome sequence datasets.

Systematically benchmarking peptide-MHC binding predictors: From synthetic to naturally processed epitopes.

A number of machine learning-based predictors have been developed for identifying immunogenic T-cell epitopes based on major histocompatibility complex (MHC) class I and II binding affinities. Rationally selecting the most appropriate tool has been complicated by the evolving training data and machine learning methods. Despite the recent advances made in generating high-quality MHC-eluted, naturally processed ligandome, the reliability of new predictors on these epitopes has yet to be evaluated. This study reports the latest benchmarking on an extensive set of MHC-binding predictors by using newly available, untested data of both synthetic and naturally processed epitopes. 32 human leukocyte antigen (HLA) class I and 24 HLA class II alleles are included in the blind test set. Artificial neural network (ANN)-based approaches demonstrated better performance than regression-based machine learning and structural modeling. Among the 18 predictors benchmarked, ANN-based mhcflurry and nn_align perform the best for MHC class I 9-mer and class II 15-mer predictions, respectively, on binding/non-binding classification (Area Under Curves = 0.911). NetMHCpan4 also demonstrated comparable predictive power. Our customization of mhcflurry to a pan-HLA predictor has achieved similar accuracy to NetMHCpan. The overall accuracy of these methods are comparable between 9-mer and 10-mer testing data. However, the top methods deliver low correlations between the predicted versus the experimental affinities for strong MHC binders. When used on naturally processed MHC-ligands, tools that have been trained on elution data (NetMHCpan4 and MixMHCpred) shows better accuracy than pure binding affinity predictor. The variability of false prediction rate is considerable among HLA types and datasets. Finally, structure-based predictor of Rosetta FlexPepDock is less optimal compared to the machine learning approaches. With our benchmarking of MHC-binding and MHC-elution predictors using a comprehensive metrics, a unbiased view for establishing best practice of T-cell epitope predictions is presented, facilitating future development of methods in immunogenomics.

Successful Identification of Head and Neck Cancer (HNC) Nodal Metastasis (NM) and Extranodal Extension (ENE) Using Deep Learning Neural Networks

Despite modern imaging modalities, radiologic detection of clinically meaningful tumor characteristics, such as NM and, particularly tumor ENE remains difficult, with reported areas under the receiver operating characteristic curve (AUC) <.70 and accuracies <70%. Emerging machine learning techniques, including deep learning, have had success with computer vision analysis. We hypothesized that computer vision with deep convolutional neural networks (DCNNs) could achieve favorable performance in identifying NM and ENE on pretreatment HNC imaging. We identified and segmented 653 lymph nodes (LN) on preoperative, diagnostic, contrast-enhanced, HNC CT scans from 270 patients with HN squamous cell carcinoma who subsequently underwent LN dissection at a single institution from 2013 - 2017. Radiographic segmentations were correlated with pathology reports and labeled as “negative”, “NM”, or “ENE.” A centralized re-review of the pathology reports documenting ENE was conducted to confirm ENE presence. A 3-dimensional DCNN was constructed using multilayer convolutional neural networks with input features consisting of normalized Hounsfield units corresponding to CT voxels within and surrounding the segmented LN. Shuffled and stratified training (64%), validation (16%), and blinded test (20%) sets were isolated a priori. The DCNN was trained using oversampling, data augmentation, dropout, regularization, and cross-validation. Performance on the blinded test set was evaluated using AUC, accuracy, sensitivity, and specificity. ENE detection performance was tested on LN samples with short axis diameter (SAD) ≥1 cm. A separate multivariable logistic regression model with bootstrapping was constructed as a benchmark comparison, using clinical characteristics and LN SAD. There were 380 negative LNs, 153 NM without ENE, and 120 NM with ENE identified and segmented. On logistic regression, clinical N-stage and greater SAD, were predictive of ENE (each P<.001) with a negative interaction between HPV-status and SAD (P<.05). Primary site (oropharynx) and SAD were predictive of NM (each P <.001). The DCNN had higher test-set performance than the benchmark logistic model for ENE (AUC: .91 vs .81, accuracy: 85.7% vs 77.7%) and NM prediction, and both models outperformed historical controls for ENE prediction (Table 1). Deep neural networks identify HNC NM and ENE with superior predictive performance than traditional logistic models and human-operator historical controls and show promise as clinician decision-making tools. External validation and prospective testing is being planned to determine the generalizability of these algorithms.Abstract 118; Table 1ENENMDCNNLogisticDCNNLogisticAUC.91.81.91.86Accuracy (%)85.777.785.576.1Sensitivity.88.72.84.79Specificity.85.80.87.74 Open table in a new tab

Pretreatment Identification of Head and Neck Cancer Nodal Metastasis and Extranodal Extension Using Deep Learning Neural Networks

Identification of nodal metastasis and tumor extranodal extension (ENE) is crucial for head and neck cancer management, but currently only can be diagnosed via postoperative pathology. Pretreatment, radiographic identification of ENE, in particular, has proven extremely difficult for clinicians, but would be greatly influential in guiding patient management. Here, we show that a deep learning convolutional neural network can be trained to identify nodal metastasis and ENE with excellent performance that surpasses what human clinicians have historically achieved. We trained a 3-dimensional convolutional neural network using a dataset of 2,875 CT-segmented lymph node samples with correlating pathology labels, cross-validated and fine-tuned on 124 samples, and conducted testing on a blinded test set of 131 samples. On the blinded test set, the model predicted ENE and nodal metastasis each with area under the receiver operating characteristic curve (AUC) of 0.91 (95%CI: 0.85–0.97). The model has the potential for use as a clinical decision-making tool to help guide head and neck cancer patient management.

On Splitting Training and Validation Set: A Comparative Study of Cross-Validation, Bootstrap and Systematic Sampling for Estimating the Generalization Performance of Supervised Learning

Model validation is the most important part of building a supervised model. For building a model with good generalization performance one must have a sensible data splitting strategy, and this is crucial for model validation. In this study, we conducted a comparative study on various reported data splitting methods. The MixSim model was employed to generate nine simulated datasets with different probabilities of mis-classification and variable sample sizes. Then partial least squares for discriminant analysis and support vector machines for classification were applied to these datasets. Data splitting methods tested included variants of cross-validation, bootstrapping, bootstrapped Latin partition, Kennard-Stone algorithm (K-S) and sample set partitioning based on joint X–Y distances algorithm (SPXY). These methods were employed to split the data into training and validation sets. The estimated generalization performances from the validation sets were then compared with the ones obtained from the blind test sets which were generated from the same distribution but were unseen by the training/validation procedure used in model construction. The results showed that the size of the data is the deciding factor for the qualities of the generalization performance estimated from the validation set. We found that there was a significant gap between the performance estimated from the validation set and the one from the test set for the all the data splitting methods employed on small datasets. Such disparity decreased when more samples were available for training/validation, and this is because the models were then moving towards approximations of the central limit theory for the simulated datasets used. We also found that having too many or too few samples in the training set had a negative effect on the estimated model performance, suggesting that it is necessary to have a good balance between the sizes of training set and validation set to have a reliable estimation of model performance. We also found that systematic sampling method such as K-S and SPXY generally had very poor estimation of the model performance, most likely due to the fact that they are designed to take the most representative samples first and thus left a rather poorly representative sample set for model performance estimation.

Abstract LB-220: An automated breast cancer detection assay for screening in the developing world

Abstract Introduction: Breast cancer incidence is rapidly increasing globally with 1.7 million new cases and 600,000 deaths due to the disease annually. Death due to breast cancer is four to six times more frequent in the developing world compared to the USA due to lack of early detection, definitive diagnosis, and limited access to treatment. Ultrasound imaging for early detection and ultrasound guided fine needle aspiration (FNA) of suspicious lesions followed by an automated, cartridge-based analysis of methylated genes could triage malignancies for faster pathology assessment and treatment, thus enabling better use of scarce resources. Experimental Procedures: To identify a panel of methylated DNA markers to enable sensitive and specific detection of malignant breast lesions, we screened 24 methylated genes (known to be frequently methylated in malignant breast tissue and unmethylated in normal breast tissue) by Quantitative Multiplex Methylation-Specific PCR (QM-MSP) analysis. Formalin-fixed paraffin-embedded (FFPE) sections of biopsies of both benign and malignant breast lesions from the USA, China and South Africa were analyzed. Samples were divided into Training and Test sets. The Training set consisted of 206 tissues [66 invasive ductal carcinoma (IDC), 30 ductal carcinoma in situ (DCIS), 99 benign breast disease (BBD) and 11 normal breast (NB)]. The gene panel selected in the Training set was examined in an independent Test set of tissues (n=204) [65 IDC, 29 DCIS, 99 BBD and 11 NB]. Further, we optimized the technical performance of an automated, prototype breast cancer detection cartridge system for quantitative assessment of gene methylation and tested it in pilot study using FNA samples from breast cancers. Results: Analysis of the tissues in the Training set (n=206) led to the selection of a panel of 10 genes highly methylated in malignant lesions with little or no methylation in benign lesions. For the 10-gene panel to achieve a sensitivity greater than 90%, a laboratory cutoff of 14.5 cumulative methylation (CM) units (out of a possible 10,000 units) was set. In the Training set, the 10-gene panel achieved a sensitivity of 90% and a specificity of 85% with receiver operating characteristic (ROC) statistics: ROC, AUC= 0.947. In a blinded Test set of tissue samples (n=204), with a laboratory cutoff of 14.5 CM units, the 10-gene panel achieved a sensitivity of 87% and a specificity of 89%, with ROC statistics: p&amp;lt;0.001, AUC= 0.936, and provided significant accuracy for breast cancers from three countries and all molecular subtypes. In a pilot study of FNA samples using the cartridge system, robust methylation in all ten genes was detected in malignant tumors. Conclusions: QM-MSP of breast lesions led to the selection of a panel of 10 genes methylated for detection of breast cancer. We have validated the technical performance of an automated, prototype cartridge system. The study reveals the potential of methylation markers to provide fast, accurate and automated cancer detection at a low cost in developing regions globally. Research Use Only. Not for diagnostic tests. A sponsored research agreement from Cepheid to Dr. Sukumar's Lab at Johns Hopkins University. Citation Format: Bradley M. Downs, Mary Jo Fackler, Claudia Mercado-Rodriguez, Ashley Cimino-Mathews, Chuang Chen, Jing-Ping Yuan, Eunice van den Berg, Leslie M. Cope, Susan C. Harvey, Syed Z. Ali, Suzana Tulac, Kriszten J. Kocmond, Edwin W. Lai, Brian Rhees, Mike Bates, Saraswati Sukumar. An automated breast cancer detection assay for screening in the developing world [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-220.

Label-Free Quantitative Proteomics Identifies Novel Plasma Biomarkers for Distinguishing Pulmonary Tuberculosis and Latent Infection.

The lack of effective differential diagnostic methods for active tuberculosis (TB) and latent infection (LTBI) is still an obstacle for TB control. Furthermore, the molecular mechanism behind the progression from LTBI to active TB has been not elucidated. Therefore, we performed label-free quantitative proteomics to identify plasma biomarkers for discriminating pulmonary TB (PTB) from LTBI. A total of 31 overlapping proteins with significant difference in expression level were identified in PTB patients (n = 15), compared with LTBI individuals (n = 15) and healthy controls (HCs, n = 15). Eight differentially expressed proteins were verified using western blot analysis, which was 100% consistent with the proteomics results. Statistically significant differences of six proteins were further validated in the PTB group compared with the LTBI and HC groups in the training set (n = 240), using ELISA. Classification and regression tree (CART) analysis was employed to determine the ideal protein combination for discriminating PTB from LTBI and HC. A diagnostic model consisting of alpha-1-antichymotrypsin (ACT), alpha-1-acid glycoprotein 1 (AGP1), and E-cadherin (CDH1) was established and presented a sensitivity of 81.2% (69/85) and a specificity of 95.2% (80/84) in discriminating PTB from LTBI, and a sensitivity of 81.2% (69/85) and a specificity of 90.1% (64/81) in discriminating PTB from HCs. Additional validation was performed by evaluating the diagnostic model in blind testing set (n = 113), which yielded a sensitivity of 75.0% (21/28) and specificity of 96.1% (25/26) in PTB vs. LTBI, 75.0% (21/28) and 92.3% (24/26) in PTB vs. HCs, and 75.0% (21/28) and 81.8% (27/33) in PTB vs. lung cancer (LC), respectively. This study obtained the plasma proteomic profiles of different M.TB infection statuses, which contribute to a better understanding of the pathogenesis involved in the transition from latent infection to TB activation and provide new potential diagnostic biomarkers for distinguishing PTB and LTBI.

Circulating Salivary miRNA hsa-miR-221 as Clinically Validated Diagnostic Marker for Hand, Foot, and Mouth Disease in Pediatric Patients

Enhancements in the diagnostic capabilities using host biomarkers are currently much needed where sensitivity and specificity issues plague the diagnosis of Hand, Foot and Mouth Disease (HFMD) in pediatrics clinical samples. We investigated miRNome profiles of HFMD saliva samples against healthy children and developed miRNA-based diagnosis models. Our 6-miRNA scoring model predicted HFMD with an overall accuracy of 85.11% in the training set and 92.86% in the blinded test set of Singapore cohort. Blinded evaluation of the model in Taiwan HFMD cases resulted in 77.08% accuracy with the 6-miRNA model and 68.75% with the 4-miRNA model. The strongest predictor of HFMD in all of the panels, hsa-miR-221 was found to be consistently and significantly downregulated in all of our HFMD cohorts. This is the first study to prove that HFMD infection could be diagnosed by circulating miRNAs in patient's saliva. Moreover, this study also serves as a stepping stone towards the future development of other infectious disease diagnosis workflows using novel biomarkers.

Four-Gene Pan-African Blood Signature Predicts Progression to Tuberculosis.

Rationale: Contacts of patients with tuberculosis (TB) constitute an important target population for preventive measures because they are at high risk of infection with Mycobacterium tuberculosis and progression to disease.Objectives: We investigated biosignatures with predictive ability for incident TB.Methods: In a case-control study nested within the Grand Challenges 6-74 longitudinal HIV-negative African cohort of exposed household contacts, we employed RNA sequencing, PCR, and the pair ratio algorithm in a training/test set approach. Overall, 79 progressors who developed TB between 3 and 24 months after diagnosis of index case and 328 matched nonprogressors who remained healthy during 24 months of follow-up were investigated.Measurements and Main Results: A four-transcript signature derived from samples in a South African and Gambian training set predicted progression up to two years before onset of disease in blinded test set samples from South Africa, the Gambia, and Ethiopia with little population-associated variability, and it was also validated in an external cohort of South African adolescents with latent M. tuberculosis infection. By contrast, published diagnostic or prognostic TB signatures were predicted in samples from some but not all three countries, indicating site-specific variability. Post hoc meta-analysis identified a single gene pair, C1QC/TRAV27 (complement C1q C-chain / T-cell receptor-α variable gene 27) that would consistently predict TB progression in household contacts from multiple African sites but not in infected adolescents without known recent exposure events.Conclusions: Collectively, we developed a simple whole blood-based PCR test to predict TB in recently exposed household contacts from diverse African populations. This test has potential for implementation in national TB contact investigation programs.

Predicting the Conformational Variability of Abl Tyrosine Kinase using Molecular Dynamics Simulations and Markov State Models.

Understanding protein conformational variability remains a challenge in drug discovery. The issue arises in protein kinases, whose multiple conformational states can affect the binding of small-molecule inhibitors. To overcome this challenge, we propose a comprehensive computational framework based on Markov state models (MSMs). Our framework integrates the information from explicit-solvent molecular dynamics simulations to accurately rank-order the accessible conformational variants of a target protein. We tested the methodology using Abl kinase with a reference and blind-test set. Only half of the Abl conformational variants discovered by our approach are present in the disclosed X-ray structures. The approach successfully identified a protein conformational state not previously observed in public structures but evident in a retrospective analysis of Lilly in-house structures: the X-ray structure of Abl with WHI-P154. Using a MSM-derived model, the free energy landscape and kinetic profile of Abl was analyzed in detail highlighting opportunities for targeting the unique metastable states.

Early Detection of Preeclampsia Using Circulating Small non-coding RNA

Preeclampsia is one of the most dangerous pregnancy complications, and the leading cause of maternal and perinatal mortality and morbidity. Although the clinical symptoms appear late, its origin is early, and hence detection is feasible already at the first trimester. In the current study, we investigated the abundance of circulating small non-coding RNAs in the plasma of pregnant women in their first trimester, seeking transcripts that best separate the preeclampsia samples from those of healthy pregnant women. To this end, we performed small non-coding RNAs sequencing of 75 preeclampsia and control samples, and identified 25 transcripts that were differentially expressed between preeclampsia and the control groups. Furthermore, we utilized those transcripts and created a pipeline for a supervised classification of preeclampsia. Our pipeline generates a logistic regression model using a 5-fold cross validation on numerous random partitions into training and blind test sets. Using this classification procedure, we achieved an average AUC value of 0.86. These findings suggest the predictive value of circulating small non-coding RNA in the first trimester, warranting further examination, and lay the foundation for producing a novel early non-invasive diagnostic tool for preeclampsia, which could reduce the life-threatening risk for both the mother and fetus.

Voltammetric fingerprinting of oils and its combination with chemometrics for the detection of extra virgin olive oil adulteration

In the present work, two approaches for the voltammetric fingerprinting of oils and their combination with chemometrics were investigated in order to detect the adulteration of extra virgin olive oil with olive pomace oil as well as the most common seed oils, namely sunflower, soybean and corn oil. In particular, cyclic voltammograms of diluted extra virgin olive oils, regular (pure) olive oils (blends of refined olive oils with virgin olive oils), olive pomace oils and seed oils in presence of dichloromethane and 0.1 M of LiClO4 in EtOH as electrolyte were recorded at a glassy carbon working electrode. Cyclic voltammetry was also employed in methanolic extracts of olive and seed oils. Datapoints of cyclic voltammograms were exported and submitted to Principal Component Analysis (PCA), Partial Least Square- Discriminant Analysis (PLS-DA) and soft independent modeling of class analogy (SIMCA). In diluted oils, PLS-DA provided a clear discrimination between olive oils (extra virgin and regular) and olive pomace/seed oils, while SIMCA showed a clear discrimination of extra virgin olive oil in regard to all other samples. Using methanolic extracts and considering datapoints recorded between 0.6 and 1.3 V, PLS-DA provided more information, resulting in three clusters-extra virgin olive oils, regular olive oils and seed/olive pomace oils-while SIMCA showed inferior performance. For the quantification of extra virgin olive oil adulteration with olive pomace oil or seed oils, a model based on Partial Least Square (PLS) analysis was developed. Detection limit of adulteration in olive oil was found to be 2% (v/v) and the linearity range up to 33% (v/v). Validation and applicability of all models was proved using a suitable test set. In the case of PLS, synthetic oil mixtures with 4 known adulteration levels in the range of 4–26% were also employed as a blind test set.

Sorting Through the Safety Data Haystack: Using Machine Learning to Identify Individual Case Safety Reports in Social-Digital Media.

IntroductionThere is increasing interest in social digital media (SDM) as a data source for pharmacovigilance activities; however, SDM is considered a low information content data source for safety data. Given that pharmacovigilance itself operates in a high-noise, lower-validity environment without objective ‘gold standards’ beyond process definitions, the introduction of large volumes of SDM into the pharmacovigilance workflow has the potential to exacerbate issues with limited manual resources to perform adverse event identification and processing. Recent advances in medical informatics have resulted in methods for developing programs which can assist human experts in the detection of valid individual case safety reports (ICSRs) within SDM.ObjectiveIn this study, we developed rule-based and machine learning (ML) models for classifying ICSRs from SDM and compared their performance with that of human pharmacovigilance experts.MethodsWe used a random sampling from a collection of 311,189 SDM posts that mentioned Roche products and brands in combination with common medical and scientific terms sourced from Twitter, Tumblr, Facebook, and a spectrum of news media blogs to develop and evaluate three iterations of an automated ICSR classifier. The ICSR classifier models consisted of sub-components to annotate the relevant ICSR elements and a component to make the final decision on the validity of the ICSR. Agreement with human pharmacovigilance experts was chosen as the preferred performance metric and was evaluated by calculating the Gwet AC1 statistic (gKappa). The best performing model was tested against the Roche global pharmacovigilance expert using a blind dataset and put through a time test of the full 311,189-post dataset.ResultsDuring this effort, the initial strict rule-based approach to ICSR classification resulted in a model with an accuracy of 65% and a gKappa of 46%. Adding an ML-based adverse event annotator improved the accuracy to 74% and gKappa to 60%. This was further improved by the addition of an additional ML ICSR detector. On a blind test set of 2500 posts, the final model demonstrated a gKappa of 78% and an accuracy of 83%. In the time test, it took the final model 48 h to complete a task that would have taken an estimated 44,000 h for human experts to perform.ConclusionThe results of this study indicate that an effective and scalable solution to the challenge of ICSR detection in SDM includes a workflow using an automated ML classifier to identify likely ICSRs for further human SME review.Electronic supplementary materialThe online version of this article (10.1007/s40264-018-0641-7) contains supplementary material, which is available to authorized users.

Antibody side chain conformations are position-dependent.

Side chain prediction is an integral component of computational antibody design and structure prediction. Current antibody modelling tools use backbone-dependent rotamer libraries with conformations taken from general proteins. Here we present our antibody-specific rotamer library, where rotamers are binned according to their immunogenetics (IMGT) position, rather than their local backbone geometry. We find that for some amino acid types at certain positions, only a restricted number of side chain conformations are ever observed. Using this information, we are able to reduce the breadth of the rotamer sampling space. Based on our rotamer library, we built a side chain predictor, position-dependent antibody rotamer swapper (PEARS). On a blind test set of 95 antibody model structures, PEARS had the highest average χ1 and χ1+2 accuracy (78.7% and 64.8%) compared to three leading backbone-dependent side chain predictors. Our use of IMGT position, rather than backbone ϕ/ψ, meant that PEARS was more robust to errors in the backbone of the model structure. PEARS also achieved the lowest number of side chain-side chain clashes. PEARS is freely available as a web application at http://opig.stats.ox.ac.uk/webapps/pears.

Identification of the Peptide PyroQ-βCasein194-209 as a Highly Specific and Sensitive Marker to Differentiate between Ultrahigh-Temperature Processed (UHT) Milk and Mildly Heated Milk.

In this study, a new approach was introduced to identify marker peptides that reflect the thermal treatment of commercial milk samples and differentiate ultrahigh-temperature processed (UHT) milk from mildly heated milk. Peptide profiles of training set samples, pasteurized (n = 20), extended shelf life (n = 29), and UHT (n = 29) milk, were recorded by MALDI-TOF-MS after StageTip microextraction. As marker candidates, 13 peptides were selected, and their cutoff levels were defined. The quality of the cutoff levels was then tested with a blind test set. Thus, the peptide m/z 1701.0, which was identified as pyroQ-βcasein194-209, could ideally differentiate UHT milk from mildly heated milk with an accuracy of 100%. Due to its high reliability and sensitivity, this peptide may be applied in routine analysis to monitor thermal processing of milk. An additional heating experiment showed that the marker peptide candidates are formed during milk processing by endogenous enzymes and selective thermal cleavage.

MALDI-TOF MS as a tool to identify foodborne yeasts and yeast-like fungi

Since food spoilage by yeasts causes high economic losses, fast and accurate identifications of yeasts associated with food and food-related products are important for the food industry. In this study the efficiency of the matrix assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) to identify food related yeasts was evaluated.A CBS in-house MALDI-TOF MS database was created and later challenged with a blinded test set of 146 yeast strains obtained from food and food related products. Ninety eight percent of the strains were correctly identified with log score values>1.7. One strain, Mrakia frigida, gained a correct identification with a score value<1.7. Two strains could not be identified at first as they represented a mix of two different species. These mixes were Rhodotorula babjevae with Meyerozyma caribbica and Clavispora lusitaniae with Debaryomyces hansenii. After separation, all four species could be correctly identified with scores>1.7. Ambiguous identifications were observed due to two incorrect reference mass spectra's found in the commercial database BDAL v.4.0, namely Candida sake DSM 70763 which was re-identified as Candida oleophila, and Candida inconspicua DSM 70631 which was re-identified as Pichia membranifaciens. MALDI-TOF MS can distinguish between most of the species, but for some species complexes, such as the Kazachstania telluris and Mrakia frigida complexes, MALDI-TOF MS showed limited resolution and identification of sibling species was sometimes problematic. Despite this, we showed that the MALDI-TOF MS is applicable for routine identification and validation of foodborne yeasts, but a further update of the commercial reference databases is needed.

Development and Validation of a High-Throughput Mass Spectrometry Based Urine Metabolomic Test for the Detection of Colonic Adenomatous Polyps.

Background: Colorectal cancer is one of the leading causes of cancer deaths worldwide. The detection and removal of the precursors to colorectal cancer, adenomatous polyps, is the key for screening. The aim of this study was to develop a clinically scalable (high throughput, low cost, and high sensitivity) mass spectrometry (MS)-based urine metabolomic test for the detection of adenomatous polyps. Methods: Prospective urine and stool samples were collected from 685 participants enrolled in a colorectal cancer screening program to undergo colonoscopy examination. Statistical analysis was performed on 69 urine metabolites measured by one-dimensional nuclear magnetic resonance spectroscopy to identify key metabolites. A targeted MS assay was then developed to quantify the key metabolites in urine. A MS-based urine metabolomic diagnostic test for adenomatous polyps was established using 67% samples (un-blinded training set) and validated using the remaining 33% samples (blinded testing set). Results: The MS-based urine metabolomic test identifies patients with colonic adenomatous polyps with an AUC of 0.692, outperforming the NMR based predictor with an AUC of 0.670. Conclusion: Here we describe a clinically scalable MS-based urine metabolomic test that identifies patients with adenomatous polyps at a higher level of sensitivity (86%) over current fecal-based tests (<18%).

DNA methylation-based forensic age prediction using artificial neural networks and next generation sequencing

The ability to estimate the age of the donor from recovered biological material at a crime scene can be of substantial value in forensic investigations. Aging can be complex and is associated with various molecular modifications in cells that accumulate over a person’s lifetime including epigenetic patterns. The aim of this study was to use age-specific DNA methylation patterns to generate an accurate model for the prediction of chronological age using data from whole blood. In total, 45 age-associated CpG sites were selected based on their reported age coefficients in a previous extensive study and investigated using publicly available methylation data obtained from 1156 whole blood samples (aged 2–90 years) analysed with Illumina’s genome-wide methylation platforms (27K/450K). Applying stepwise regression for variable selection, 23 of these CpG sites were identified that could significantly contribute to age prediction modelling and multiple regression analysis carried out with these markers provided an accurate prediction of age (R2=0.92, mean absolute error (MAE)=4.6 years). However, applying machine learning, and more specifically a generalised regression neural network model, the age prediction significantly improved (R2=0.96) with a MAE=3.3 years for the training set and 4.4 years for a blind test set of 231 cases. The machine learning approach used 16 CpG sites, located in 16 different genomic regions, with the top 3 predictors of age belonged to the genes NHLRC1, SCGN and CSNK1D. The proposed model was further tested using independent cohorts of 53 monozygotic twins (MAE=7.1 years) and a cohort of 1011 disease state individuals (MAE=7.2 years). Furthermore, we highlighted the age markers’ potential applicability in samples other than blood by predicting age with similar accuracy in 265 saliva samples (R2=0.96) with a MAE=3.2 years (training set) and 4.0 years (blind test). In an attempt to create a sensitive and accurate age prediction test, a next generation sequencing (NGS)-based method able to quantify the methylation status of the selected 16 CpG sites was developed using the Illumina MiSeq® platform. The method was validated using DNA standards of known methylation levels and the age prediction accuracy has been initially assessed in a set of 46 whole blood samples. Although the resulted prediction accuracy using the NGS data was lower compared to the original model (MAE=7.5years), it is expected that future optimization of our strategy to account for technical variation as well as increasing the sample size will improve both the prediction accuracy and reproducibility.

Rapid Discrimination between Methicillin-Sensitive and Methicillin-Resistant Staphylococcus aureus Using MALDI-TOF Mass Spectrometry.

Methicillin-resistant Staphylococcus aureus (MRSA) is one of the major pathogens responsible for nosocomial infections. The presence of MRSA in a hospital is detrimental to patients and to hospital management. Thus, rapid identification of MRSA is needed. Here, we report on a prospective method to rapidly discriminate of MSSA from MRSA using matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) and support vector machine (SVM) analysis in 160 clinical isolates of S. aureus. The predictive model was tested using 100 S. aureus isolates (50 MSSA and 50 MRSA). The identification rates were 90.0% for MSSA and 87.5% for MRSA in a 10-fold cross-validation SVM. In blind test sets, 60 S. aureus isolates (30 MSSA and 30 MRSA) were correctly classified, with identification rates of 93.3% for MSSA and 86.7% for MRSA. The method proposed in this study using the predictive model enables detection of one colony in 5 minutes, and thus is useful at clinical sites at which rapid discrimination of MRSA from MSSA is required.