Invasive or More Direct Measurements Can Provide an Objective Early-Stopping Ceiling for Training Deep Neural Networks on Non-invasive or Less-Direct Biomedical Data.

Early stopping is an extremely common tool to minimize overfitting, which would otherwise be a cause of poor generalization of the model to novel data. However, early stopping is a heuristic that, while effective, primarily relies on ad hoc parameters and metrics. Optimizing when to stop remains a challenge. In this paper, we suggest that for some biomedical applications, a natural dichotomy of invasive/non-invasive measurements, or more generally proximal vs distal measurements of a biological system can be exploited to provide objective advice on early stopping. We discuss the conditions where invasive measurements of a biological process should provide better predictions than non-invasive measurements, or at best offer parity. Hence, if data from an invasive measurement are available locally, or from the literature, that information can be leveraged to know with high certainty whether a model of non-invasive data is overfitted. We present paired invasive/non-invasive cardiac and coronary artery measurements from two mouse strains, one of which spontaneously develops type 2 diabetes, posed as a classification problem. Examination of the various stopping rules shows that generalization is reduced with more training epochs and commonly applied stopping rules give widely different generalization error estimates. The use of an empirically derived training ceiling is demonstrated to be helpful as added information to leverage early stopping in order to reduce overfitting.

The Importance of Weakly Co-Evolving Residue Networks in Proteins is Revealed by Visual Analytics.

Small changes in a protein's core packing produce changes in function, and even small changes in function bias species fitness and survival. Therefore individually deleterious mutations should be evolutionarily coupled with compensating mutations that recover fitness. Co-evolving pairs of mutations should be littered across evolutionary history. Despite longstanding intuition, the results of co-evolution analyses have largely disappointed expectations. Regardless of the statistics applied, only a small majority of the most strongly co-evolving residues are typically found to be in contact, and much of the "meaning" of observed co-evolution has been opaque. In a medium-sized protein of 300 amino acids, there are almost 20 million potentially-important interdependencies. It is impossible to understand this data in textual format without extreme summarization or truncation. And, due to summarization and truncation, it is impossible to identify most patterns in the data. We developed a visualization approach that eschews the common "look at a long list of statistics" approach and instead enables the user to literally look at all of the co-evolution statistics simultaneously. Users of our tool reported visually obvious "clouds" of co-evolution statistics forming distinct patterns in the data, and analysis demonstrated that these clouds had structural relevance. To determine whether this phenomenon generalized, we repeated this experiment in three proteins we had not previously studied. The results provide evidence about how structural constrains have impacted co-evolution, why previous "examine the most frequently co-evolving residues" approaches have had limited success, and additionally shed light on the biophysical importance of different types of co-evolution.

Chromatin structure in cancer.

In the past decade, we have seen the emergence of sequence-based methods to understand chromosome organization. With the confluence of in situ approaches to capture information on looping, topological domains, and larger chromatin compartments, understanding chromatin-driven disease is becoming feasible. Excitingly, recent advances in single molecule imaging with capacity to reconstruct "bulk-cell" features of chromosome conformation have revealed cell-to-cell chromatin structural variation. The fundamental question motivating our analysis of the literature is, can altered chromatin structure drive tumorigenesis? As our community learns more about rare disease, including low mutational frequency cancers, understanding "chromatin-driven" pathology will illuminate the regulatory structures of the genome. We describe recent insights into altered genome architecture in human cancer, highlighting multiple pathways toward disruptions of chromatin structure, including structural variation, noncoding mutations, metabolism, and de novo mutations to architectural regulators themselves. Our analysis of the literature reveals that deregulation of genome structure is characteristic in distinct classes of chromatin-driven tumors. As we begin to integrate the findings from single cell imaging studies and chromatin structural sequencing, we will be able to understand the diversity of cells within a common diagnosis, and begin to define structure-function relationships of the misfolded genome.

Improvement of automated analysis of coronary Doppler echocardiograms.

Coronary artery disease is the leading cause of heart disease, and while it can be assessed through transthoracic Doppler echocardiography (TTDE) by observing changes in coronary flow, manual analysis of TTDE is time consuming and subject to bias. In a previous study, a program was created to automatically analyze coronary flow patterns by parsing Doppler videos into a single continuous image, binarizing and separating the image into cardiac cycles, and extracting data values from each of these cycles. The program significantly reduced variability and time to complete TTDE analysis, but some obstacles such as interfering noise and varying video sizes left room to increase the program's accuracy. The goal of this current study was to refine the existing automation algorithm and heuristics by (1) moving the program to a Python environment, (2) increasing the program's ability to handle challenging cases and video variations, and (3) removing unrepresentative cardiac cycles from the final data set. With this improved analysis, examiners can use the automatic program to easily and accurately identify the early signs of serious heart diseases.

A 1% TBSA Chart Reduces Math Errors While Retaining Acceptable First-Estimate Accuracy.

Life-threatening and treatment-altering errors occur in estimates of the percentage of total body surface area burned (%TBSA burned) with unacceptable frequency. In response, numerous attempts have been made to improve the charts commonly used for %TBSA-burned estimation. Recent research shows that the largest errors in %TBSA-burned estimates probably come from sources other than inaccurate values in the charts. Here, we develop a taxonomy of the possible sources of error and their impact on %TBSA-burned estimates. Also, we observe that different caregivers have different estimation needs: First-responders require a rapid estimate with sufficient accuracy to enable them to begin care and determine patient transport options, while burn surgeons ordering skin grafts desire accuracy to the square centimeter, and can afford considerable time to attain that accuracy. These competing needs suggest that a one-tool-fits-all-caregivers approach is suboptimal. We therefore present a validated, simplified burn chart that minimizes one of the largest sources of random errors in %TBSA-burned estimates-simple calculation errors-while also being quick and requiring little training. NCHart-1 also enables simple consensus estimates, as well as separation of estimation subtasks across caregivers, leading to several potential improvements in mass casualty situations. Our results demonstrate that NCHart-1 possesses the accuracy necessary for first responders, while reliably producing results in less than 2 minutes. Of 76 healthcare professionals surveyed, a large majority indicated a preference for NCHart-1 over their previous methods for ease of both use and training. For clinical or commercial use of NCHart-1, please contact: tech.commercialization@nationwidechildrens.org.

Forming Big Datasets through Latent Class Concatenation of Imperfectly Matched Databases Features.

Informatics researchers often need to combine data from many different sources to increase statistical power and study subtle or complicated effects. Perfect overlap of measurements across academic studies is rare since virtually every dataset is collected for a unique purpose and without coordination across parties not-at-hand (i.e., informatics researchers in the future). Thus, incomplete concordance of measurements across datasets poses a major challenge for researchers seeking to combine public databases. In any given field, some measurements are fairly standard, but every organization collecting data makes unique decisions on instruments, protocols, and methods of processing the data. This typically denies literal concatenation of the raw data since constituent cohorts do not have the same measurements (i.e., columns of data). When measurements across datasets are similar prima facie, there is a desire to combine the data to increase power, but mixing non-identical measurements could greatly reduce the sensitivity of the downstream analysis. Here, we discuss a statistical method that is applicable when certain patterns of missing data are found; namely, it is possible to combine datasets that measure the same underlying constructs (or latent traits) when there is only partial overlap of measurements across the constituent datasets. Our method, ROSETTA empirically derives a set of common latent trait metrics for each related measurement domain using a novel variation of factor analysis to ensure equivalence across the constituent datasets. The advantage of combining datasets this way is the simplicity, statistical power, and modeling flexibility of a single joint analysis of all the data. Three simulation studies show the performance of ROSETTA on datasets with only partially overlapping measurements (i.e., systematically missing information), benchmarked to a condition of perfectly overlapped data (i.e., full information). The first study examined a range of correlations, while the second study was modeled after the observed correlations in a well-characterized clinical, behavioral cohort. Both studies consistently show significant correlations >0.94, often >0.96, indicating the robustness of the method and validating the general approach. The third study varied within and between domain correlations and compared ROSETTA to multiple imputation and meta-analysis as two commonly used methods that ostensibly solve the same data integration problem. We provide one alternative to meta-analysis and multiple imputation by developing a method that statistically equates similar but distinct manifest metrics into a set of empirically derived metrics that can be used for analysis across all datasets.

In Silico Modeling of Biofilm Formation by Nontypeable Haemophilus influenzae In Vivo.

Biofilms formed by nontypeable Haemophilus influenzae (NTHI) bacteria play an important role in multiple respiratory tract diseases. Visual inspection of the morphology of biofilms formed during chronic infections shows distinct differences from biofilms formed in vitro To better understand these differences, we analyzed images of NTHI biofilms formed in the middle ears of Chinchilla lanigera and developed an in silico agent-based model of the formation of NTHI biofilms in vivo We found that, as in vitro, NTHI bacteria are organized in self-similar patterns; however, the sizes of NTHI clusters in vivo are more than 10-fold smaller than their in vitro counterparts. The agent-based model reproduced these patterns and suggested that smaller clusters occur due to elimination of planktonic NTHI cells by the host responses. Estimation of model parameters by fitting simulation results to imaging data showed that the effects of several processes in the model change during the course of the infection.IMPORTANCE Multiple respiratory illnesses are associated with formation of biofilms within the human airway by NTHI. However, a substantial amount of our understanding of the mechanisms that underlie NTHI biofilm formation is obtained from in vitro studies. Our in silico model that describes biofilm formation by NTHI within the middle ears of Chinchilla lanigera will help isolate processes potentially responsible for the differences between the morphologies of biofilms formed in vivo versus those formed in vitro Thus, the in silico model can be used to glean mechanisms that underlie biofilm formation in vivo and connect those mechanisms to those obtained from in vitro experiments. The in silico model developed here can be extended to investigate potential roles of specific host responses (e.g., mucociliary clearance) on NTHI biofilm formation in vivo The developed computational tools can also be used to analyze and describe biofilm formation by other bacterial species in vivo.

Comparison of the Lund and Browder table to computed tomography scan three-dimensional surface area measurement for a pediatric cohort.

BACKGROUND: Treating burns effectively requires accurately assessing the percentage of the total body surface area (%TBSA) affected by burns. Current methods for estimating %TBSA, such as Lund and Browder (L&B) tables, rely on historic body statistics. An increasingly obese population has been blamed for increasing errors in %TBSA estimates. However, this assumption has not been experimentally validated. We hypothesized that errors in %TBSA estimates using L&B were due to differences in the physical proportions of today's children compared with children in the early 1940s when the chart was developed and that these differences would appear as body mass index (BMI)-associated systematic errors in the L&B values versus actual body surface areas. MATERIALS AND METHODS: We measured the TBSA of human pediatric cadavers using computed tomography scans. Subjects ranged from 9 mo to 15 y in age. We chose outliers of the BMI distribution (from the 31st percentile at the low through the 99th percentile at the high). We examined surface area proportions corresponding to L&B regions. RESULTS: Measured regional proportions based on computed tomography scans were in reasonable agreement with L&B, even with subjects in the tails of the BMI range. The largest deviation was 3.4%, significantly less than the error seen in real-world %TBSA estimates. CONCLUSIONS: While today's population is more obese than those studied by L&B, their body region proportions scale surprisingly well. The primary error in %TBSA estimation is not due to changing physical proportions of today's children and may instead lie in the application of the L&B table.

Extracellular DNA and Type IV Pilus Expression Regulate the Structure and Kinetics of Biofilm Formation by Nontypeable Haemophilus influenzae.

Biofilms formed in the middle ear by nontypeable Haemophilus influenzae (NTHI) are central to the chronicity, recurrence, and refractive nature of otitis media (OM). However, mechanisms that underlie the emergence of specific NTHI biofilm structures are unclear. We combined computational analysis tools and in silico modeling rooted in statistical physics with confocal imaging of NTHI biofilms formed in vitro during static culture in order to identify mechanisms that give rise to distinguishing morphological features. Our analysis of confocal images of biofilms formed by NTHI strain 86-028NP using pair correlations of local bacterial densities within sequential planes parallel to the substrate showed the presence of fractal structures of short length scales (≤10 µm). The in silico modeling revealed that extracellular DNA (eDNA) and type IV pilus (Tfp) expression played important roles in giving rise to the fractal structures and allowed us to predict a substantial reduction of these structures for an isogenic mutant (ΔcomE) that was significantly compromised in its ability to release eDNA into the biofilm matrix and had impaired Tfp function. This prediction was confirmed by analysis of confocal images of in vitro ΔcomE strain biofilms. The fractal structures potentially generate niches for NTHI survival in the hostile middle ear microenvironment by dramatically increasing the contact area of the biofilm with the surrounding environment, facilitating nutrient exchange, and by generating spatial positive feedback to quorum signaling.IMPORTANCE NTHI is a major bacterial pathogen for OM, which is a common ear infection in children worldwide. Chronic OM is associated with bacterial biofilm formation in the middle ear; therefore, knowledge of the mechanisms that underlie NTHI biofilm formation is important for the development of therapeutic strategies for NTHI-associated OM. Our combined approach using confocal imaging of NTHI biofilms formed in vitro and mathematical tools for analysis of pairwise density correlations and agent-based modeling revealed that eDNA and Tfp expression were important factors in the development of fractal structures in NTHI biofilms. These structures may help NTHI survive in hostile environments, such as the middle ear. Our in silico model can be used in combination with laboratory or animal modeling studies to further define the mechanisms that underlie NTHI biofilm development during OM and thereby guide the rational design of, and optimize time and cost for, benchwork and preclinical studies.

Facilitating surgeon understanding of complex anatomy using a three-dimensional printed model.

BACKGROUND: 3-dimensional prints (3DP) anecdotally facilitate surgeon understanding of anatomy and decision-making. However, the actual benefit to surgeons or patients has not been quantified. This study investigates how surgeon understanding of complex anatomy is altered by a 3DP compared to computed tomography (CT) scan or CT + digital reconstruction (CT + DR). MATERIALS AND METHODS: Key anatomic features were segmented from a CT-abdomen/pelvis of pygopagus twins to build a DR and printed in color on a 3D printer. Pediatric surgery trainees and attendings (n = 21) were tested regarding anatomy identification and their understanding of point-to-point distances, scale, and shape. RESULTS: There was no difference between media regarding point-to-point distances. The 3DP led to an increased number of correct answers for questions of scale and shape compared to CT (P < 0.05). CT + DR performance was intermediate but not statistically different from 3DP or CT. Identification of anatomy was inconsistent between media; however, answers were significantly closer to correct when using the 3DP. Participants completed the test faster with the 3DP (6.6 ± 0.5 min) (P < 0.05) than with CT (18.9 ± 2.5 min) or CT + 3DR (14.9 ± 1.5 min). CONCLUSIONS: Although point-to-point measurements were not different, 3DP increased the understanding of shape, scale, and anatomy. It enabled understanding significantly faster than other media. In difficult surgical cases with complex anatomy and a need for efficient multidisciplinary coordination, 3D printed models should be considered for surgical planning.

In silico modeling identifies CD45 as a regulator of IL-2 synergy in the NKG2D-mediated activation of immature human NK cells.

Natural killer (NK) cells perform immunosurveillance of virally infected and transformed cells, and their activation depends on the balance between signaling by inhibitory and activating receptors. Cytokine receptor signaling can synergize with activating receptor signaling to induce NK cell activation. We investigated the interplay between the signaling pathways stimulated by the cytokine interleukin-2 (IL-2) and the activating receptor NKG2D in immature (CD56bright) and mature (CD56dim) subsets of human primary NK cells using mass cytometry experiments and in silico modeling. Our analysis revealed that IL-2 changed the abundances of several key proteins, including NKG2D and the phosphatase CD45. Furthermore, we found differences in correlations between protein abundances, which were associated with the maturation state of the NK cells. The mass cytometry measurements also revealed that the signaling kinetics of key protein abundances induced by NKG2D stimulation depended on the maturation state and the pretreatment condition of the NK cells. Our in silico model, which described the multidimensional data with coupled first-order reactions, predicted that the increase in CD45 abundance was a major enhancer of NKG2D-mediated activation in IL-2-treated CD56bright NK cells but not in IL-2-treated CD56dim NK cells. This dependence on CD45 was verified by measurement of CD107a mobilization to the NK cell surface (a marker of activation). Our mathematical framework can be used to glean mechanisms underlying synergistic signaling pathways in other activated immune cells.

Investigating skin-to-skin care patterns with extremely preterm infants in the NICU and their effect on early cognitive and communication performance: a retrospective cohort study.

OBJECTIVES: The primary objective of the study was to investigate how patterns of skin-to-skin care might impact infant early cognitive and communication performance. DESIGN: This was a retrospective cohort study. SETTING: This study took place in a level-IV all-referral neonatal intensive care unit in the Midwest USA specialising in the care of extremely preterm infants. PARTICIPANTS: Data were collected from the electronic medical records of all extremely preterm infants (gestational age <27 weeks) admitted to the unit during 2010-2011 and who completed 6-month and 12-month developmental assessments in the follow-up clinic (n=97). OUTCOME MEASURES: Outcome measures included the cognitive and communication subscales of the Bayley Scales of Infant Development, Third Edition (Bayley-III); and skin-to-skin patterns including: total hours of maternal and paternal participation throughout hospitalisation, total duration in weeks and frequency (hours per week). ANALYSIS: Extracted data were analysed through a multistep process of logistic regressions, t-tests, χ2 tests and Fisher's exact tests followed with exploratory network analysis using novel visual analytic software. RESULTS: Infants who received above the sample median in total hours, weekly frequency and total hours from mothers and fathers of skin-to-skin care were more likely to score ≥80 on the cognitive and communication scales of the Bayley-III. However, the results were not statistically significant (p>0.05). Mothers provided the majority of skin-to-skin care with a sharp decline at 30 weeks corrected age, regardless of when extremely preterm infants were admitted. Additional exploratory network analysis suggests that medical and skin-to-skin factors play a parallel, non-synergistic role in contributing to early cognitive and communication performance as assessed through the Bayley-III. CONCLUSIONS: This study suggests an association between early and frequent skin-to-skin care with extremely preterm infants and early cognitive and communication performance.

Valve Endothelial Cell-Derived Tgfß1 Signaling Promotes Nuclear Localization of Sox9 in Interstitial Cells Associated With Attenuated Calcification.

OBJECTIVE: Aortic valve disease, including calcification, affects >2% of the human population and is caused by complex interactions between multiple risk factors, including genetic mutations, the environment, and biomechanics. At present, there are no effective treatments other than surgery, and this is because of the limited understanding of the mechanisms that underlie the condition. Previous work has shown that valve interstitial cells within the aortic valve cusps differentiate toward an osteoblast-like cell and deposit bone-like matrix that leads to leaflet stiffening and calcific aortic valve stenosis. However, the mechanisms that promote pathological phenotypes in valve interstitial cells are unknown. APPROACH AND RESULTS: Using a combination of in vitro and in vivo tools with mouse, porcine, and human tissue, we show that in valve interstitial cells, reduced Sox9 expression and nuclear localization precedes the onset of calcification. In vitro, Sox9 nuclear export and calcific nodule formation is prevented by valve endothelial cells. However, in vivo, loss of Tgfß1 in the endothelium leads to reduced Sox9 expression and calcific aortic valve disease. CONCLUSIONS: Together, these findings suggest that reduced nuclear localization of Sox9 in valve interstitial cells is an early indicator of calcification, and therefore, pharmacological targeting to prevent nuclear export could serve as a novel therapeutic tool in the prevention of calcification and stenosis.

MoFlow: visualizing conformational changes in molecules as molecular flow improves understanding.

BACKGROUND: Current visualizations of molecular motion use a Timeline-analogous representation that conveys "first the molecule was shaped like this, then like this...". This scheme is orthogonal to the Pathline-like human understanding of motion "this part of the molecule moved from here to here along this path". We present MoFlow, a system for visualizing molecular motion using a Pathline-analogous representation. RESULTS: The MoFlow system produces high-quality renderings of molecular motion as atom pathlines, as well as interactive WebGL visualizations, and 3D printable models. In a preliminary user study, MoFlow representations are shown to be superior to canonical representations for conveying molecular motion. CONCLUSIONS: Pathline-based representations of molecular motion are more easily understood than timeline representations. Pathline representations provide other advantages because they represent motion directly, rather than representing structure with inferred motion.

Optimization of Synthetic Proteins: Identification of Interpositional Dependencies Indicating Structurally and/or Functionally Linked Residues.

Protein alignments are commonly used to evaluate the similarity of protein residues, and the derived consensus sequence used for identifying functional units (e.g., domains). Traditional consensus-building models fail to account for interpositional dependencies - functionally required covariation of residues that tend to appear simultaneously throughout evolution and across the phylogentic tree. These relationships can reveal important clues about the processes of protein folding, thermostability, and the formation of functional sites, which in turn can be used to inform the engineering of synthetic proteins. Unfortunately, these relationships essentially form sub-motifs which cannot be predicted by simple "majority rule" or even HMM-based consensus models, and the result can be a biologically invalid "consensus" which is not only never seen in nature but is less viable than any extant protein. We have developed a visual analytics tool, StickWRLD, which creates an interactive 3D representation of a protein alignment and clearly displays covarying residues. The user has the ability to pan and zoom, as well as dynamically change the statistical threshold underlying the identification of covariants. StickWRLD has previously been successfully used to identify functionally-required covarying residues in proteins such as Adenylate Kinase and in DNA sequences such as endonuclease target sites.

Rational design and engineering of a modified adeno-associated virus (AAV1)-based vector system for enhanced retrograde gene delivery.

BACKGROUND: After injection into muscle and peripheral nerves, a variety of viral vectors undergo retrograde transport to lower motor neurons. However, because of its attractive safety profile and durable gene expression, adeno-associated virus (AAV) remains the only vector to have been applied to the human nervous system for the treatment of neurodegenerative disease. Nonetheless, only a very small fraction of intramuscularly injected AAV vector arrives at the spinal cord. OBJECTIVE: To engineer a novel AAV vector by inserting a neuronal targeting peptide (Tet1), with binding properties similar to those of tetanus toxin, into the AAV1 capsid. METHODS: Integral to this approach was the use of structure-based design to increase the effectiveness of functional capsid engineering. This approach allowed the optimization of scaffolding regions for effective display of the foreign epitope while minimizing disruption of the native capsid structure. We also validated an approach by which low-titer tropism-modified AAV vectors can be rescued by particle mosaicism with unmodified capsid proteins. RESULTS: Importantly, our rationally engineered AAV1-based vectors exhibited markedly enhanced transduction of cultured motor neurons, diminished transduction of nontarget cells, and markedly superior retrograde delivery compared with unmodified AAV1 vector. CONCLUSION: This approach promises a significant advancement in the rational engineering of AAV vectors for diseases of the nervous system and other organs.

StickWRLD as an Interactive Visual Pre-Filter for Canceromics-Centric Expression Quantitative Trait Locus Data.

As datasets increase in complexity, the time required for analysis (both computational and human domain-expert) increases. One of the significant impediments introduced by such burgeoning data is the difficulty in knowing what features to include or exclude from statistical models. Simple tables of summary statistics rarely provide an adequate picture of the patterns and details of the dataset to enable researchers to make well-informed decisions about the adequacy of the models they are constructing. We have developed a tool, StickWRLD, which allows the user to visually browse through their data, displaying all possible correlations. By allowing the user to dynamically modify the retention parameters (both P and the residual, r), StickWRLD allows the user to identify significant correlations and disregard potential correlations that do not meet those same criteria - effectively filtering through all possible correlations quickly and identifying possible relationships of interest for further analysis. In this study, we applied StickWRLD to a semi-synthetic dataset constructed from two published human datasets. In addition to detecting high-probability correlations in this dataset, we were able to quickly identify gene-SNP correlations that would have gone undetected using more traditional approaches due to issues of low penetrance.

Collection and extraction of saliva DNA for next generation sequencing.

The preferred source of DNA in human genetics research is blood, or cell lines derived from blood, as these sources yield large quantities of high quality DNA. However, DNA extraction from saliva can yield high quality DNA with little to no degradation/fragmentation that is suitable for a variety of DNA assays without the expense of a phlebotomist and can even be acquired through the mail. However, at present, no saliva DNA collection/extraction protocols for next generation sequencing have been presented in the literature. This protocol optimizes parameters of saliva collection/storage and DNA extraction to be of sufficient quality and quantity for DNA assays with the highest standards, including microarray genotyping and next generation sequencing.

Understanding the sequence requirements of protein families: insights from the BioVis 2013 contests.

INTRODUCTION: In 2011, the BioVis symposium of the IEEE VisWeek conferences inaugurated a new variety of data analysis contest. Aimed at fostering collaborations between computational scientists and biologists, the BioVis contest provided real data from biological domains with emerging visualization needs, in the hope that novel approaches would result in powerful new tools for the community. In 2011 and 2012 the theme of these contests was expression Quantitative Trait Locus analysis, within and across tissues respectively. In 2013 the topic was updated to protein sequence and mutation visualization. METHODS: The contest was framed in the context of a real protein with numerous mutations that had lost function, and the question posed "what minimal set of changes would you propose to rescue function, or how could you support a biologist attempting to answer that question?". The data was grounded in actual experimental results in triosephosphate isomerase(TIM) enzymes. Seven teams composed of 36 individuals submitted entries with proposed solutions and approaches to the challenge. Their contributions ranged from careful analysis of the visualization and analytical requirements for the problem through integration of existing tools for analyzing the context and consequences of protein mutations, to completely new tools addressing the problem. RESULTS: Judges found valuable and novel contributions in each of the entries, including interesting ways to hierarchicalize the protein into domains of informational interaction, tools for simultaneously understanding both sequential and spatial order, and approaches for conveying some types of inter-residue dependencies. In this manuscript we document the problem presented to the contestants, summarize the biological contributions of their entries, and suggest opportunities that this work has highlighted for even more improved tools in the future.

Addressing the unmet need for visualizing conditional random fields in biological data.

BACKGROUND: The biological world is replete with phenomena that appear to be ideally modeled and analyzed by one archetypal statistical framework - the Graphical Probabilistic Model (GPM). The structure of GPMs is a uniquely good match for biological problems that range from aligning sequences to modeling the genome-to-phenome relationship. The fundamental questions that GPMs address involve making decisions based on a complex web of interacting factors. Unfortunately, while GPMs ideally fit many questions in biology, they are not an easy solution to apply. Building a GPM is not a simple task for an end user. Moreover, applying GPMs is also impeded by the insidious fact that the "complex web of interacting factors" inherent to a problem might be easy to define and also intractable to compute upon. DISCUSSION: We propose that the visualization sciences can contribute to many domains of the bio-sciences, by developing tools to address archetypal representation and user interaction issues in GPMs, and in particular a variety of GPM called a Conditional Random Field(CRF). CRFs bring additional power, and additional complexity, because the CRF dependency network can be conditioned on the query data. CONCLUSIONS: In this manuscript we examine the shared features of several biological problems that are amenable to modeling with CRFs, highlight the challenges that existing visualization and visual analytics paradigms induce for these data, and document an experimental solution called StickWRLD which, while leaving room for improvement, has been successfully applied in several biological research projects. Software and tutorials are available at http://www.stickwrld.org/.