Big Data to Knowledge: Application of Machine Learning to Predictive Modeling of Therapeutic Response in Cancer

BackgroundIn recent years, the availability of high throughput technologies, establishment of large molecular patient data repositories, and advancement in computing power and storage have allowed elucidation of complex mechanisms implicated in therapeutic response in cancer patients. The breadth and depth of such data, alongside experimental noise and missing values, requires a sophisticated human-machine interaction that would allow effective learning from complex data and accurate forecasting of future outcomes, ideally embedded in the core of machine learning design.ObjectiveIn this review, we will discuss machine learning techniques utilized for modeling of treatment response in cancer, including Random Forests, support vector machines, neural networks, and linear and logistic regression. We will overview their mathematical foundations and discuss their limitations and alternative approaches in light of their application to therapeutic response modeling in cancer.ConclusionWe hypothesize that the increase in the number of patient profiles and potential temporal monitoring of patient data will define even more complex techniques, such as deep learning and causal analysis, as central players in therapeutic response modeling.     

From Gigabyte to Kilobyte: A Bioinformatics Protocol for Mining Large RNA-Seq Transcriptomics Data

RNA-Seq techniques generate hundreds of millions of short RNA reads using next-generation sequencing (NGS). These RNA reads can be mapped to reference genomes to investigate changes of gene expression but improved procedures for mining large RNA-Seq datasets to extract valuable biological knowledge are needed. RNAMiner—a multi-level bioinformatics protocol and pipeline—has been developed for such datasets. It includes five steps: Mapping RNA-Seq reads to a reference genome, calculating gene expression values, identifying differentially expressed genes, predicting gene functions, and constructing gene regulatory networks. To demonstrate its utility, we applied RNAMiner to datasets generated from Human, Mouse, Arabidopsis thaliana, and Drosophila melanogaster cells, and successfully identified differentially expressed genes, clustered them into cohesive functional groups, and constructed novel gene regulatory networks. The RNAMiner web service is available at http://calla.rnet.missouri.edu/rnaminer/index.html.     

2009 Society for Medical Anthropology Conference: From Race to Repositories: Bioethics in a Genomics Age

Medical anthropologist Barbara Koenig spoke on the intersection of bioethics and genomics at the 2009 Society for Medical Anthropology Conference at Yale University in New Haven, Connecticut.Medical anthropologist Barbara Koenig is at the forefront of numerous endeavors that seek to understand race in the age of genomics and protect human subjects with regard to DNA repositories. Through a dialogue ranging from the history and critiques of bioethics to her own research projects, Koenig brought her respect and excitement for the anthropology found in biomedicine to the 2009 Society for Medical Anthropology conference at Yale University in September.As a whole, the space in which bioethics operates does not constitute a “field” in the traditional sense. It is an amalgam of approaches to the social, ethical, and cultural issues surrounding biology and medicine. While the prevalent medical ethic for 2,500 years has been understood to be embodied in the Hippocratic Oath [1], the modern age of biomedical research and clinical practice demands more careful inspection of ethical issues in biology and medicine. This precise space is where bioethics has found fertile ground for investigation. For Koenig, anthropological analysis and, more specifically, empirical ethnographic study are critical tools to tease apart the complexities of bioethical issues.Recently, Koenig, a professor of medicine at the Mayo Clinic, has considered the definitions of race in light of the emerging work of genomics [2,3]. Her book, Revisiting Race in a Genomic Age, argues that the complex definitions of race are an interesting blend of biological differences and cultural traditions. Often in modern society, the unity assumed among a racial group is at odds with the underlying biological diversity revealed by genomics. And yet individuals in our species are much more related to one another than, for example, any two chimps are related. It is clear from the much-cited work of the Human Genome Project [4,5] and the International HapMap Consortium [6] that genomics will have a deeply profound impact on how we understand our ancestry and how easily individuals will be able to trace their own origins using modern genetic technologies.Hints at new implications in personalized medicine also have come to light, showing that different patient populations, distinguished by race, can have significantly different responses to drugs [7,8]. Increasingly, race is used as a categorization to parse individuals for the purposes of biomedical research. These categories are themselves areas of contention, as they are socially constructed. Individuals may identify themselves with different racial categories depending on the context in which the identifiers are provided. Furthermore, it is unclear what implications can be drawn from the studies in which researchers are asked to conclude something about a social identity based on biological data. Research in these areas undoubtedly will provide insightful anthropological analysis in the coming years.Much of bioethics is concerned with the issues surrounding new medical technologies and their impact on society. In this same vein, the advent of sequencing databases and direct-to-consumer genetic tests has had broad implications for the participants. At the Mayo Clinic, Koenig is working on the development of a centralized biobank. This facility will store DNA samples collected from patients with a wide range of diseases and enable easy access for clinicians and scientists interested in studying a given disease. However, as with any collection of information, there is a question of access. Who can see what information? How will this be regulated? And to what extent is informed consent required for each of the future uses of the collected sample?There has been much public debate regarding the ethical use of forensic DNA databases [9,10] and yet less public consideration concerning the medical DNA databases. Koenig spoke of attempts to create community understanding beyond the informed consent standard. An auxiliary issue related to the social impacts of new genetic methods is the emergence of direct-to-consumer genetic testing. Currently, certain companies market to the public with the promise of assessing the customer’s genetic profile in order to identify genetic risk factors for disease. Without the intermediary of a health care provider to validate the analysis and contextualize the risk, these tests can have an alarming and bewildering effect on consumers. How should one understand an 8 percent increase in colorectal cancer risk relative to the general population? Certainly, these new technologies and the companies pushing them directly to the consumer public are of profound anthropological interest.While some anthropologists analyze trends and conduct surveys to understand their subjects, Koenig has entrenched herself within the clinic and on advisory boards to shape the outcome of bioethical issues facing our society in a time of ever-changing technologies. She has done so with a firm foundation in the historical perspective and practical ethic of biomedicine.     

Predictive Outcome Modeling of Preoperative Clinical Symptoms and Electrodiagnostic Data in Tarsal Tunnel Surgery

Background  The relationship between tarsal tunnel syndrome (TTS), electrodiagnostic (Edx) findings, and surgical outcome is unknown. Analysis of TTS surgical release outcome patient satisfaction and comparison to Edx nerve conduction studies (NCSs) is important to improve outcome prediction when deciding who would benefit from TTS release. Methods  Retrospective study of 90 patients over 7 years that had tarsal tunnel (TT) release surgery with outcome rating and preoperative tibial NCS. Overall, 64 patients met study inclusion criteria with enough NCS data to be classified into one of the following three groups: (1) probable TTS, (2) peripheral polyneuropathy, or (3) normal. Most patients had preoperative clinical provocative testing including diagnostic tibial nerve injection, tibial Phalen''s sign, and/or Tinel''s sign and complaints of plantar tibial neuropathic symptoms. Outcome measure was percentage of patient improvement report at surgical follow-up visit. Results  Patient-reported improvement was 92% in the probable TTS group ( n  = 41) and 77% of the non-TTS group ( n  = 23). Multivariate modeling revealed that three out of eight variables predicted improvement from surgical release, NCS consistent with TTS ( p  = 0.04), neuropathic symptoms ( p  = 0.045), and absent Phalen''s test ( p  = 0.001). The R ² was 0.21 which is a robust result for this outcome measurement process. Conclusion  The best predictors of improvement in patients with TTS release were found in patients that had preoperative Edx evidence of tibial neuropathy in the TT and tibial nerve plantar symptoms. Determining what factors predict surgical outcome will require prospective evaluation and evaluation of patients with other nonsurgical modalities.     

Predictive Modeling of In Vivo Response to Gemcitabine in Pancreatic Cancer

A clear contradiction exists between cytotoxic in-vitro studies demonstrating effectiveness of Gemcitabine to curtail pancreatic cancer and in-vivo studies failing to show Gemcitabine as an effective treatment. The outcome of chemotherapy in metastatic stages, where surgery is no longer viable, shows a 5-year survival <5%. It is apparent that in-vitro experiments, no matter how well designed, may fail to adequately represent the complex in-vivo microenvironmental and phenotypic characteristics of the cancer, including cell proliferation and apoptosis. We evaluate in-vitro cytotoxic data as an indicator of in-vivo treatment success using a mathematical model of tumor growth based on a dimensionless formulation describing tumor biology. Inputs to the model are obtained under optimal drug exposure conditions in-vitro. The model incorporates heterogeneous cell proliferation and death caused by spatial diffusion gradients of oxygen/nutrients due to inefficient vascularization and abundant stroma, and thus is able to simulate the effect of the microenvironment as a barrier to effective nutrient and drug delivery. Analysis of the mathematical model indicates the pancreatic tumors to be mostly resistant to Gemcitabine treatment in-vivo. The model results are confirmed with experiments in live mice, which indicate uninhibited tumor proliferation and metastasis with Gemcitabine treatment. By extracting mathematical model parameter values for proliferation and death from monolayer in-vitro cytotoxicity experiments with pancreatic cancer cells, and simulating the effects of spatial diffusion, we use the model to predict the drug response in-vivo, beyond what would have been expected from sole consideration of the cancer intrinsic resistance. We conclude that this integrated experimental/computational approach may enhance understanding of pancreatic cancer behavior and its response to various chemotherapies, and, further, that such an approach could predict resistance based on pharmacokinetic measurements with the goal to maximize effective treatment strategies.     

Mechanical-Statistical Modeling in Ecology: From Outbreak Detections to Pest Dynamics

Knowledge about large-scale and long-term dynamics of (natural) populations is required to assess the efficiency of control strategies, the potential for long-term persistence, and the adaptability to global changes such as habitat fragmentation and global warming. For most natural populations, such as pest populations, large-scale and long-term surveys cannot be carried out at a high resolution. For instance, for population dynamics characterized by irregular abundance explosions, i.e., outbreaks, it is common to report detected outbreaks rather than measuring the population density at every location and time event. Here, we propose a mechanical-statistical model for analyzing such outbreak occurrence data and making inference about population dynamics. This spatio-temporal model contains the main mechanisms of the dynamics and describes the observation process. This construction enables us to account for the discrepancy between the phenomenon scale and the sampling scale. We propose the Bayesian method to estimate model parameters, pest densities and hidden factors, i.e., variables involved in the dynamics but not observed. The model was specified and used to learn about the dynamics of the European pine sawfly (Neodiprion sertifer Geoffr., an insect causing major defoliation of pines in northern Europe) based on Finnish sawfly data covering the years 1961–1990. In this application, a dynamical Beverton–Holt model including a hidden regime variable was incorporated into the model to deal with large variations in the population densities. Our results gave support to the idea that pine sawfly dynamics should be studied as metapopulations with alternative equilibria. The results confirmed the importance of extreme minimum winter temperatures for the occurrence of European pine sawfly outbreaks. The strong positive connection between the ratio of lake area over total area and outbreaks was quantified for the first time.     

Maintenance Integration in Manufacturing Systems: From the Modeling Tool to Evaluation

In manufacturing systems, wear-out and eventual failure are unavoidable. However, to reduce the rate of their occurrence and to prolong the life of equipment or the capacity for extended productive use of the equipment under the necessary technological functioning and servicing, maintenance can be performed. For large manufacturing systems, maintenance integration involves a particular development concerned with both complexity models and computing time. This paper presents an effective way of modeling complex manufacturing systems through hierarchical and modular analysis by using stochastic Petri nets and Markov chains. In the proposed approach, the integration of maintenance policies in a manufacturing system is facilitated by the development of a generic model. With this generic modeling, the user doesn't need to code the strategies but only to instantiate the generic model with the structure of the manufacturing system. This method allows various maintenance strategies to be coded in the generic model with the aim of studying their influence on system dependability and performance.     

Measuring Fit of Sequence Data to Phylogenetic Model: Gain of Power Using Marginal Tests

Testing fit of data to model is fundamentally important to any science, but publications in the field of phylogenetics rarely do this. Such analyses discard fundamental aspects of science as prescribed by Karl Popper. Indeed, not without cause, Popper (Unended quest: an intellectual autobiography. Fontana, London, 1976) once argued that evolutionary biology was unscientific as its hypotheses were untestable. Here we trace developments in assessing fit from Penny et al. (Nature 297:197–200, 1982) to the present. We compare the general log-likelihood ratio (the G or G ² statistic) statistic between the evolutionary tree model and the multinomial model with that of marginalized tests applied to an alignment (using placental mammal coding sequence data). It is seen that the most general test does not reject the fit of data to model (P ~ 0.5), but the marginalized tests do. Tests on pairwise frequency (F) matrices, strongly (P < 0.001) reject the most general phylogenetic (GTR) models commonly in use. It is also clear (P < 0.01) that the sequences are not stationary in their nucleotide composition. Deviations from stationarity and homogeneity seem to be unevenly distributed amongst taxa; not necessarily those expected from examining other regions of the genome. By marginalizing the 4 ^t patterns of the i.i.d. model to observed and expected parsimony counts, that is, from constant sites, to singletons, to parsimony informative characters of a minimum possible length, then the likelihood ratio test regains power, and it too rejects the evolutionary model with P ≪ 0.001. Given such behavior over relatively recent evolutionary time, readers in general should maintain a healthy skepticism of results, as the scale of the systematic errors in published trees may really be far larger than the analytical methods (e.g., bootstrap) report.     

From Sedentary Time to Sedentary Patterns: Accelerometer Data Reduction Decisions in Youth

Aim

This study aims to establish evidence-based accelerometer data reduction criteria to accurately assess total sedentary time and sedentary patterns in children.

Methods

Participants (n = 1057 European children; 9–13 yrs) were invited to wear an accelerometer for at least 6 consecutive days. We explored 1) non-wear time criteria; 2) minimum daily valid wear time; 3) differences between weekday and weekend day; and 4) minimum number of days of accelerometer wear by comparing the effects of commonly used data reduction decisions on total sedentary time, and duration and number of prolonged sedentary bouts.

Results

More than 60 consecutive minutes of zero counts was the optimal criterion for non-wear time. Increasing the definition of a valid day from 8 to 10 hours wear time hardly influenced the sedentary outcomes, while the sample size of children with more than 4 valid days increased from 69 to 81%. On weekdays, children had on average 1 hour more wear time, 50 minutes more total sedentary time, 26 minutes more sedentary time accumulated in bouts, and 1 more sedentary bout. At least 6 days of accelerometer data were needed to accurately represent weekly sedentary time and patterns.

Conclusions

Based on our results we recommend 1) a minimum of 60 minutes of consecutive zeros as the most realistic criterion for non-wear time; and 2) including at least six days with minimum eight valid hours to characterize children''s usual total sedentary time and patterns, preferably including one weekend day.     

Sequence errors described in GenBank: a means to determine the accuracy of DNA sequence interpretation.

The accuracy of nucleic acid sequence data interpretation was determined by assessing and quantifying the discrepancies reported in the GenBank database. This permitted the calculation of an Error Rate (ER) for nucleic acid sequence determination. If one assumes that most entries (TB, Total Bases) were independently verified or those without reported discrepancies were correct, the ER is 0.368 errors per 1000 bases. However, if one assumes that only those sequences with reported discrepancies (TBIQ, Total Bases from entries In Question) were verified and are thus correct, the ER is 2.887 errors per 1000 bases. This establishes the first set of limit boundaries of the ER for sequence interpretation and sequence errors within the GenBank database and provides the foundation for future assessments and the monitoring of sequence data accumulation. In addition, the ER measure provides a basis to evaluate the efficiency and merit of present and future automated nucleic acid sequencing technologies which will have a direct impact upon the final outcome of the "Human Genome Initiative".     

Activity Response to Climate Seasonality in Species with Fossorial Habits: A Niche Modeling Approach Using the Lowland Burrowing Treefrog (Smilisca fodiens)

The importance of climatic conditions in shaping the geographic distribution of amphibian species is mainly associated to their high sensitivity to environmental conditions. How they cope with climate gradients through behavioral adaptations throughout their distribution is an important issue due to the ecological and evolutionary implications for population viability. Given their low dispersal abilities, the response to seasonal climate changes may not be migration, but behavioral and physiological adaptations. Here we tested whether shifts in climatic seasonality can predict the temporal variation of surface activity of the fossorial Lowland Burrowing Treefrog (Smilisca fodiens) across its geographical distribution. We employed Ecological Niche Modeling (ENM) to perform a monthly analysis of spatial variation of suitable climatic conditions (defined by the July conditions, the month of greatest activity), and then evaluated the geographical correspondence of monthly projections with the occurrence data per month. We found that the species activity, based on the species'' occurrence data, corresponds with the latitudinal variation of suitable climatic conditions. Due to the behavioral response of this fossorial frog to seasonal climate variation, we suggest that precipitation and temperature have played a major role in the definition of geographical and temporal distribution patterns, as well as in shaping behavioral adaptations to local climatic conditions. This highlights the influence of macroclimate on shaping activity patterns and the important role of fossorials habits to meet the environmental requirements necessary for survival.     

fastSTRUCTURE: Variational Inference of Population Structure in Large SNP Data Sets

Tools for estimating population structure from genetic data are now used in a wide variety of applications in population genetics. However, inferring population structure in large modern data sets imposes severe computational challenges. Here, we develop efficient algorithms for approximate inference of the model underlying the STRUCTURE program using a variational Bayesian framework. Variational methods pose the problem of computing relevant posterior distributions as an optimization problem, allowing us to build on recent advances in optimization theory to develop fast inference tools. In addition, we propose useful heuristic scores to identify the number of populations represented in a data set and a new hierarchical prior to detect weak population structure in the data. We test the variational algorithms on simulated data and illustrate using genotype data from the CEPH–Human Genome Diversity Panel. The variational algorithms are almost two orders of magnitude faster than STRUCTURE and achieve accuracies comparable to those of ADMIXTURE. Furthermore, our results show that the heuristic scores for choosing model complexity provide a reasonable range of values for the number of populations represented in the data, with minimal bias toward detecting structure when it is very weak. Our algorithm, fastSTRUCTURE, is freely available online at http://pritchardlab.stanford.edu/structure.html.     

How to Determine the Accuracy of an Alternative Diagnostic Test when It Is Actually Better than the Reference Tests: A Re-Evaluation of Diagnostic Tests for Scrub Typhus Using Bayesian LCMs

BackgroundThe indirect immunofluorescence assay (IFA) is considered a reference test for scrub typhus. Recently, the Scrub Typhus Infection Criteria (STIC; a combination of culture, PCR assays and IFA IgM) were proposed as a reference standard for evaluating alternative diagnostic tests. Here, we use Bayesian latent class models (LCMs) to estimate the true accuracy of each diagnostic test, and of STIC, for diagnosing scrub typhus.ConclusionsThe low specificity of STIC was caused by the low specificity of IFA IgM. Neither STIC nor IFA IgM can be used as reference standards against which to evaluate alternative diagnostic tests. Further evaluation of new diagnostic tests should be done with a carefully selected set of diagnostic tests and appropriate statistical models.