Identification of a serum proteomic biomarker panel using diagnosis specific ensemble learning and symptoms for early pancreatic cancer detection.

BACKGROUND: The grim (<10% 5-year) survival rates for pancreatic ductal adenocarcinoma (PDAC) are attributed to its complex intrinsic biology and most often late-stage detection. The overlap of symptoms with benign gastrointestinal conditions in early stage further complicates timely detection. The suboptimal diagnostic performance of carbohydrate antigen (CA) 19-9 and elevation in benign hyperbilirubinaemia undermine its reliability, leaving a notable absence of accurate diagnostic biomarkers. Using a selected patient cohort with benign pancreatic and biliary tract conditions we aimed to develop a data analysis protocol leading to a biomarker signature capable of distinguishing patients with non-specific yet concerning clinical presentations, from those with PDAC. METHODS: 539 patient serum samples collected under the Accelerated Diagnosis of neuro Endocrine and Pancreatic TumourS (ADEPTS) study (benign disease controls and PDACs) and the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS, healthy controls) were screened using the Olink Oncology II panel, supplemented with five in-house markers. 16 specialized base-learner classifiers were stacked to select and enhance biomarker performances and robustness in blinded samples. Each base-learner was constructed through cross-validation and recursive feature elimination in a discovery set comprising approximately two thirds of the ADEPTS and UKCTOCS samples and contrasted specific diagnosis with PDAC. RESULTS: The signature which was developed using diagnosis-specific ensemble learning demonstrated predictive capabilities outperforming CA19-9, the only biomarker currently accepted by the FDA and the National Comprehensive Cancer Network guidelines for pancreatic cancer, and other individual biomarkers and combinations in both discovery and held-out validation sets. An AUC of 0.98 (95% CI 0.98-0.99) and sensitivity of 0.99 (95% CI 0.98-1) at 90% specificity was achieved with the ensemble method, which was significantly larger than the AUC of 0.79 (95% CI 0.66-0.91) and sensitivity 0.67 (95% CI 0.50-0.83), also at 90% specificity, for CA19-9, in the discovery set (p = 0.0016 and p = 0.00050, respectively). During ensemble signature validation in the held-out set, an AUC of 0.95 (95% CI 0.91-0.99), sensitivity 0.86 (95% CI 0.68-1), was attained compared to an AUC of 0.80 (95% CI 0.66-0.93), sensitivity 0.65 (95% CI 0.48-0.56) at 90% specificity for CA19-9 alone (p = 0.0082 and p = 0.024, respectively). When validated only on the benign disease controls and PDACs collected from ADEPTS, the diagnostic-specific signature achieved an AUC of 0.96 (95% CI 0.92-0.99), sensitivity 0.82 (95% CI 0.64-0.95) at 90% specificity, which was still significantly higher than the performance for CA19-9 taken as a single predictor, AUC of 0.79 (95% CI 0.64-0.93) and sensitivity of 0.18 (95% CI 0.03-0.69) (p = 0.013 and p = 0.0055, respectively). CONCLUSION: Our ensemble modelling technique outperformed CA19-9, individual biomarkers and indices developed with prevailing algorithms in distinguishing patients with non-specific but concerning symptoms from those with PDAC, with implications for improving its early detection in individuals at risk.

A novel framework for inferring parameters of transmission from viral sequence data.

Transmission between hosts is a critical part of the viral lifecycle. Recent studies of viral transmission have used genome sequence data to evaluate the number of particles transmitted between hosts, and the role of selection as it operates during the transmission process. However, the interpretation of sequence data describing transmission events is a challenging task. We here present a novel and comprehensive framework for using short-read sequence data to understand viral transmission events, designed for influenza virus, but adaptable to other viral species. Our approach solves multiple shortcomings of previous methods for this purpose; for example, we consider transmission as an event involving whole viruses, rather than sets of independent alleles. We demonstrate how selection during transmission and noisy sequence data may each affect naive inferences of the population bottleneck, accounting for these in our framework so as to achieve a correct inference. We identify circumstances in which selection for increased viral transmission may or may not be identified from data. Applying our method to experimental data in which transmission occurs in the presence of strong selection, we show that our framework grants a more quantitative insight into transmission events than previous approaches, inferring the bottleneck in a manner that accounts for selection, both for within-host virulence, and for inherent viral transmissibility. Our work provides new opportunities for studying transmission processes in influenza, and by extension, in other infectious diseases.

Association between the cervicovaginal microbiome, BRCA1 mutation status, and risk of ovarian cancer: a case-control study.

BACKGROUND: Various factors-including age, family history, inflammation, reproductive factors, and tubal ligation-modulate the risk of ovarian cancer. In this study, our aim was to establish whether women with, or at risk of developing, ovarian cancer have an imbalanced cervicovaginal microbiome. METHODS: We did a case-control study in two sets of women aged 18-87 years in the Czech Republic, Germany, Italy, Norway, and the UK. The ovarian cancer set comprised women with epithelial ovarian cancer and controls (both healthy controls and those diagnosed with benign gynaecological conditions). The BRCA set comprised women with a BRCA1 mutation but without ovarian cancer and controls who were wild type for BRCA1 and BRCA2 (both healthy controls and those with benign gynaecological conditions). Cervicovaginal samples were gathered from all participants with the ThinPrep system and then underwent 16S rRNA gene sequencing. For each sample, we calculated the proportion of lactobacilli species (ie, Lactobacillus crispatus, Lactobacillus iners, Lactobacillus gasseri, and Lactobacillus jensenii), which are essential for the generation of a protective low vaginal pH, in the cervicovaginal microbiota. We grouped samples into those in which lactobacilli accounted for at least 50% of the species present (community type L) and those in which lactobacilli accounted for less than 50% of the species present (community type O). We assessed the adjusted association between BRCA1 status and ovarian cancer status and cervicovaginal microbiota community type, using a logistic regression model with a bias reduction method. FINDINGS: Participants were recruited between Jan 2, 2016, and July 21, 2018. The ovarian cancer set (n=360) comprised 176 women with epithelial ovarian cancer, 115 healthy controls and 69 controls with benign gynaecological conditions. The BRCA set (n=220) included 109 women with BRCA1 mutations, 97 healthy controls wild type for BRCA1 and BRCA2 and 14 controls with a benign gynaecological condition wild type for BRCA1 and BRCA2. On the basis of two-dimensional density plots, receiver-operating characteristic curve analysis, and age thresholds used previously, we divided the cohort into those younger than 50 years and those aged 50 years or older. In the ovarian cancer set, women aged 50 years or older had a higher prevalence of community type O microbiota (81 [61%] of 133 ovarian cancer cases and 84 [59%] of 142 healthy controls) than those younger than 50 years (23 [53%] of 43 cases and 12 [29%] of 42 controls). In the ovarian cancer set, women younger than 50 years with ovarian cancer had a significantly higher prevalence of community type O microbiota than did age-matched controls under a logistic regression model with bias correction (odds ratio [OR] 2·80 [95% CI 1·17-6·94]; p=0·020). In the BRCA set, women with BRCA1 mutations younger than 50 years were also more likely to have community type O microbiota than age-matched controls (OR 2·79 [95% CI 1·25-6·68]; p=0·012), after adjustment for pregnancy (ever). This risk was increased further if more than one first-degree family member was affected by any cancer (OR 5·26 [95% CI 1·83-15·30]; p=0·0022). In both sets, we noted that the younger the participants, the stronger the association between community type O microbiota and ovarian cancer or BRCA1 mutation status (eg, OR for community type O for cases aged <40 years in the ovarian cancer set 7·00 [95% CI 1·27-51·44], p=0·025; OR for community type O for BRCA1 mutation carriers aged <35 years in the BRCA set 4·40 [1·14-24·36], p=0·031). INTERPRETATION: The presence of ovarian cancer, or factors known to affect risk for the disease (ie, age and BRCA1 germline mutations), were significantly associated with having a community type O cervicovaginal microbiota. Whether re-instatement of a community type L microbiome by using, for example, vaginal suppositories containing live lactobacilli, would alter the microbiomial composition higher up in the female genital tract and in the fallopian tubes (the site of origin of high-grade serous ovarian cancer), and whether such changes could translate into a reduced incidence of ovarian cancer, needs to be investigated. FUNDING: EU Horizon 2020 Research and Innovation Programme, EU Horizon 2020 European Research Council Programme, and The Eve Appeal.

Serum biomarker-based early detection of pancreatic ductal adenocarcinomas with ensemble learning.

BACKGROUND: Earlier detection of pancreatic ductal adenocarcinoma (PDAC) is key to improving patient outcomes, as it is mostly detected at advanced stages which are associated with poor survival. Developing non-invasive blood tests for early detection would be an important breakthrough. METHODS: The primary objective of the work presented here is to use a dataset that is prospectively collected, to quantify a set of cancer-associated proteins and construct multi-marker models with the capacity to predict PDAC years before diagnosis. The data used is part of a nested case-control study within the UK Collaborative Trial of Ovarian Cancer Screening and is comprised of 218 samples, collected from a total of 143 post-menopausal women who were diagnosed with pancreatic cancer within 70 months after sample collection, and 249 matched non-cancer controls. We develop a stacked ensemble modelling technique to achieve robustness in predictions and, therefore, improve performance in newly collected datasets. RESULTS: Here we show that with ensemble learning we can predict PDAC status with an AUC of 0.91 (95% CI 0.75-1.0), sensitivity of 92% (95% CI 0.54-1.0) at 90% specificity, up to 1 year prior to diagnosis, and at an AUC of 0.85 (95% CI 0.74-0.93) up to 2 years prior to diagnosis (sensitivity of 61%, 95% CI 0.17-0.83, at 90% specificity). CONCLUSIONS: The ensemble modelling strategy explored here outperforms considerably biomarker combinations cited in the literature. Further developments in the selection of classifiers balancing performance and heterogeneity should further enhance the predictive capacity of the method.

Pancreatic cancers are most frequently detected at an advanced stage. This limits treatment options and contributes to the dismal survival rates currently recorded. The development of new tests that could improve detection of early-stage disease is fundamental to improve outcomes. Here, we use advanced data analysis techniques to devise an early detection test for pancreatic cancer. We use data on markers in the blood from people enrolled on a screening trial. Our test correctly identifies as positive for pancreatic cancer 91% of the time up to 1 year prior to diagnosis, and 78% of the time up to 2 years prior to diagnosis. These results surpass previously reported tests and should encourage further evaluation of the test in different populations, to see whether it should be adopted in the clinic.

DNA methylation signatures to predict the cervicovaginal microbiome status.

BACKGROUND: The composition of the microbiome plays an important role in human health and disease. Whether there is a direct association between the cervicovaginal microbiome and the host's epigenome is largely unexplored. RESULTS: Here we analyzed a total of 448 cervicovaginal smear samples and studied both the DNA methylome of the host and the microbiome using the Illumina EPIC array and next-generation sequencing, respectively. We found that those CpGs that are hypo-methylated in samples with non-lactobacilli (O-type) dominating communities are strongly associated with gastrointestinal differentiation and that a signature consisting of 819 CpGs was able to discriminate lactobacilli-dominating (L-type) from O-type samples with an area under the receiver operator characteristic curve (AUC) of 0.84 (95% CI = 0.77-0.90) in an independent validation set. The performance found in samples with more than 50% epithelial cells was further improved (AUC 0.87) and in women younger than 50 years of age was even higher (AUC 0.91). In a subset of 96 women, the buccal but not the blood cell DNA showed the same trend as the cervicovaginal samples in discriminating women with L- from O-type cervicovaginal communities. CONCLUSIONS: These findings strongly support the view that the epithelial epigenome plays an essential role in hosting specific microbial communities.

Sensitivity of asymmetric rate-dependent critical systems to initial conditions: Insights into cellular decision making.

The work reported here aims to address the effects of time-dependent parameters and stochasticity on decision making in biological systems. We achieve this by extending previous studies that resorted to simple bifurcation normal forms, although in the present case we focus primarily on the issue of the system's sensitivity to initial conditions in the presence of two different noise distributions, Gaussian and Lévy. In addition, we also assess the impact of two-way sweeping at different rates through the critical region of a canonical Pitchfork bifurcation with a constant external asymmetry. The parallel with decision making in biocircuits is performed on this simple system since it is equivalent in its available states and dynamics to more complex genetic circuits published previously. Overall we verify that rate-dependent effects, previously reported as being important features of bifurcating systems, are specific to particular initial conditions. Processing of each starting state, which for the normal form underlying this study is akin to a classification task, is affected by the balance between sweeping speed through critical regions and the type of fluctuations added. For the heavy-tailed noise, two-way dynamic bifurcations are more efficient in processing the external signals, here understood to be jointly represented by the critical parameter profile and the external asymmetry amplitude, when compared to the system relying on escape dynamics. This is particular to the case when the system starts at an attractor not favored by the asymmetry and, in conjunction, when the sweeping amplitude is large.

Inferring Fitness Effects from Time-Resolved Sequence Data with a Delay-Deterministic Model.

A common challenge arising from the observation of an evolutionary system over time is to infer the magnitude of selection acting upon a specific genetic variant, or variants, within the population. The inference of selection may be confounded by the effects of genetic drift in a system, leading to the development of inference procedures to account for these effects. However, recent work has suggested that deterministic models of evolution may be effective in capturing the effects of selection even under complex models of demography, suggesting the more general application of deterministic approaches to inference. Responding to this literature, we here note a case in which a deterministic model of evolution may give highly misleading inferences, resulting from the nondeterministic properties of mutation in a finite population. We propose an alternative approach that acts to correct for this error, and which we denote the delay-deterministic model. Applying our model to a simple evolutionary system, we demonstrate its performance in quantifying the extent of selection acting within that system. We further consider the application of our model to sequence data from an evolutionary experiment. We outline scenarios in which our model may produce improved results for the inference of selection, noting that such situations can be easily identified via the use of a regular deterministic model.

Decision making in noisy bistable systems with time-dependent asymmetry.

Our work draws special attention to the importance of the effects of time-dependent parameters on decision making in bistable systems. Here, we extend previous studies of the mechanism known as speed-dependent cellular decision making in genetic circuits by performing an analytical treatment of the canonical supercritical pitchfork bifurcation problem with an additional time-dependent asymmetry and control parameter. This model has an analogous behavior to the genetic switch. In the presence of transient asymmetries and fluctuations, slow passage through the critical region in both systems increases substantially the probability of specific decision outcomes. We also study the relevance for attractor selection of reaching maximum values for the external asymmetry before and after the critical region. Overall, maximum asymmetries should be reached at an instant where the position of the critical point allows for compensation of the detrimental effects of noise in retaining memory of the transient asymmetries.

Interplay between path and speed in decision making by high-dimensional stochastic gene regulatory networks.

Induction of a specific transcriptional program by external signaling inputs is a crucial aspect of intracellular network functioning. The theoretical concept of coexisting attractors representing particular genetic programs is reasonably adapted to experimental observations of "genome-wide" expression profiles or phenotypes. Attractors can be associated either with developmental outcomes such as differentiation into specific types of cells, or maintenance of cell functioning such as proliferation or apoptosis. Here we review a mechanism known as speed-dependent cellular decision making (SdCDM) in a small epigenetic switch and generalize the concept to high-dimensional space. We demonstrate that high-dimensional network clustering capacity is dependent on the level of intrinsic noise and the speed at which external signals operate on the transcriptional landscape.

Speed-dependent cellular decision making in nonequilibrium genetic circuits.

Despite being governed by the principles of nonequilibrium transitions, gene expression dynamics underlying cell fate decision is poorly understood. In particular, the effect of signaling speed on cellular decision making is still unclear. Here we show that the decision between alternative cell fates, in a structurally symmetric circuit, can be biased depending on the speed at which the system is forced to go through the decision point. The circuit consists of two mutually inhibiting and self-activating genes, forced by two external signals with identical stationary values but different transient times. Under these conditions, slow passage through the decision point leads to a consistently biased decision due to the transient signaling asymmetry, whereas fast passage reduces and eventually eliminates the switch imbalance. The effect is robust to noise and shows that dynamic bifurcations, well known in nonequilibrium physics, are important for the control of genetic circuits.