Search

The collection of immunoglobulin genes in an individual’s germline, which gives rise to B cell receptors via recombination, is known to vary significantly across individuals. In humans, for example, each individual has only a fraction of the several hundred known V alleles. Furthermore, the currently-accepted set of known V alleles is both incomplete (particularly for non-European samples), and contains a significant number of spurious alleles. The resulting uncertainty as to which immunoglobulin alleles are present in any given sample results in inaccurate B cell receptor sequence annotations, and in particular inaccurate inferred naive ancestors. In this paper we first show that the currently widespread practice of aligning each sequence to its closest match in the full set of IMGT alleles results in a very large number of spurious alleles that are not in the sample’s true set of germline V alleles. We then describe a new method for inferring each individual’s germline gene set from deep sequencing data, and show that it improves upon existing methods by making a detailed comparison on a variety of simulated and real data samples. This new method has been integrated into the partis annotation and clonal family inference package, available at https://github.com/psathyrella/partis, and is run by default without affecting overall run time., Author summary Antibodies are an important component of the adaptive immune system, which itself determines our response to both pathogens and vaccines. They are produced by B cells through somatic recombination of germline DNA, which results in a vast diversity of antigen binding affinities across the B cell repertoire. We typically learn about the development of this repertoire, and its history of interaction with antigens, by sequencing large numbers of the DNA sequences from which antibodies are derived. In order to understand such data, it is necessary to determine the combination of germline V, D, and J genes that was rearranged to form each such B cell receptor sequence. This is difficult, however, because the immunoglobulin locus exhibits an extraordinary level of diversity across individuals—encompassing both allelic variation and gene duplication, deletion, and conversion—and because the locus’s large size and repetitive structure make germline sequencing very difficult. In this paper we describe a new computational method that avoids this difficulty by inferring each individual’s set of immunoglobulin germline genes directly from expressed B cell receptor sequence data.

Phylogenetic networks are rooted, directed, acyclic graphs that model reticulate evolutionary histories. Recently, statistical methods were devised for inferring such networks from either gene tree estimates or the sequence alignments of multiple unlinked loci. Bi-allelic markers, most notably single nucleotide polymorphisms (SNPs) and amplified fragment length polymorphisms (AFLPs), provide a powerful source of genome-wide data. In a recent paper, a method called SNAPP was introduced for statistical inference of species trees from unlinked bi-allelic markers. The generative process assumed by the method combined both a model of evolution for the bi-allelic markers, as well as the multispecies coalescent. A novel component of the method was a polynomial-time algorithm for exact computation of the likelihood of a fixed species tree via integration over all possible gene trees for a given marker. Here we report on a method for Bayesian inference of phylogenetic networks from bi-allelic markers. Our method significantly extends the algorithm for exact computation of phylogenetic network likelihood via integration over all possible gene trees. Unlike the case of species trees, the algorithm is no longer polynomial-time on all instances of phylogenetic networks. Furthermore, the method utilizes a reversible-jump MCMC technique to sample the posterior of phylogenetic networks given bi-allelic marker data. Our method has a very good performance in terms of accuracy and robustness as we demonstrate on simulated data, as well as a data set of multiple New Zealand species of the plant genus Ourisia (Plantaginaceae). We implemented the method in the publicly available, open-source PhyloNet software package., Author summary The availability of genomic data has revolutionized the study of evolutionary histories and phylogeny inference. Inferring evolutionary histories from genomic data requires, in most cases, accounting for the fact that different genomic regions could have evolutionary histories that differ from each other as well as from that of the species from which the genomes were sampled. In this paper, we introduce a method for inferring evolutionary histories while accounting for two processes that could give rise to such differences across the genomes, namely incomplete lineage sorting and hybridization. We introduce a novel algorithm for computing the likelihood of phylogenetic networks from bi-allelic genetic markers and use it in a Bayesian inference method. Analyses of synthetic and empirical data sets show a very good performance of the method in terms of the estimates it obtains.

Cancer cells share several metabolic traits, including aerobic production of lactate from glucose (Warburg effect), extensive glutamine utilization and impaired mitochondrial electron flow. It is still unclear how these metabolic rearrangements, which may involve different molecular events in different cells, contribute to a selective advantage for cancer cell proliferation. To ascertain which metabolic pathways are used to convert glucose and glutamine to balanced energy and biomass production, we performed systematic constraint-based simulations of a model of human central metabolism. Sampling of the feasible flux space allowed us to obtain a large number of randomly mutated cells simulated at different glutamine and glucose uptake rates. We observed that, in the limited subset of proliferating cells, most displayed fermentation of glucose to lactate in the presence of oxygen. At high utilization rates of glutamine, oxidative utilization of glucose was decreased, while the production of lactate from glutamine was enhanced. This emergent phenotype was observed only when the available carbon exceeded the amount that could be fully oxidized by the available oxygen. Under the latter conditions, standard Flux Balance Analysis indicated that: this metabolic pattern is optimal to maximize biomass and ATP production; it requires the activity of a branched TCA cycle, in which glutamine-dependent reductive carboxylation cooperates to the production of lipids and proteins; it is sustained by a variety of redox-controlled metabolic reactions. In a K-ras transformed cell line we experimentally assessed glutamine-induced metabolic changes. We validated computational results through an extension of Flux Balance Analysis that allows prediction of metabolite variations. Taken together these findings offer new understanding of the logic of the metabolic reprogramming that underlies cancer cell growth., Author summary Hallmarks describing common key events in initiation, maintenance and progression of cancer have been identified. One hallmark deals with rewiring of metabolic reactions required to sustain enhanced cell proliferation. The availability of molecular, mechanistic models of cancer hallmarks will mightily improve optimized personal treatment and new drug discovery. Metabolism is the only hallmark for which it is currently possible to derive large scale mathematical models, which have predictive ability. In this paper, we exploit a constraint-based model of the core metabolism required for biomass conversion of the most relevant nutrients—glucose and glutamine—to clarify the logic of control of cancer metabolism. We newly report that, when available oxygen is not sufficient to fully oxidize available glucose and glutamine carbons–a situation compatible with that observed under normal oxygen conditions in human and in cancer cells growing in vitro—utilization of glutamine by reductive carboxylation and conversion of glucose and glutamine to lactate confer advantage for biomass production. Redox homeostasis can be maintained through the use of different alternative pathways. In conclusion, this paper offers a logic interpretation to the link between metabolic rewiring and enhanced proliferation, which may offer new approaches to targeted drug discovery and utilization.

Metaproteomic studies adopt the common bottom-up proteomics approach to investigate the protein composition and the dynamics of protein expression in microbial communities. When matched metagenomic and/or metatranscriptomic data of the microbial communities are available, metaproteomic data analyses often employ a metagenome-guided approach, in which complete or fragmental protein-coding genes are first directly predicted from metagenomic (and/or metatranscriptomic) sequences or from their assemblies, and the resulting protein sequences are then used as the reference database for peptide/protein identification from MS/MS spectra. This approach is often limited because protein coding genes predicted from metagenomes are incomplete and fragmental. In this paper, we present a graph-centric approach to improving metagenome-guided peptide and protein identification in metaproteomics. Our method exploits the de Bruijn graph structure reported by metagenome assembly algorithms to generate a comprehensive database of protein sequences encoded in the community. We tested our method using several public metaproteomic datasets with matched metagenomic and metatranscriptomic sequencing data acquired from complex microbial communities in a biological wastewater treatment plant. The results showed that many more peptides and proteins can be identified when assembly graphs were utilized, improving the characterization of the proteins expressed in the microbial communities. The additional proteins we identified contribute to the characterization of important pathways such as those involved in degradation of chemical hazards. Our tools are released as open-source software on github at https://github.com/COL-IU/Graph2Pro., Author Summary In recent years, meta-omic (including metatranscriptomic and metaproteomic) techniques have been adopted as complementary approaches to metagenomic sequencing to study functional characteristics and dynamics of microbial communities, aiming at a holistic understanding of a community to respond to the changes in the environment. Currently, metaproteomic data are largely analyzed using the bioinformatics tools originally designed in bottom-up proteomics. In particular, recent metaproteomic studies employed a metagenome-guided approach, in which complete or fragmental protein-coding genes were first predicted from metagenomic sequences (i.e., contigs or scaffolds), acquired from the matched community samples, and predicted protein sequences were then used in peptide identification. A key challenge of this approach is that the protein coding genes predicted from assembled metagenomic contigs can be incomplete and fragmented due to the complexity of metagenomic samples and the short reads length in metagenomic sequencing. To address this issue, in this paper, we present a graph-centric approach that exploits the de bruijn graph structure reported by metagenome assembly algorithms to improve metagenome-guided peptide and protein identification in metaproteomics. We show that our method can identify much more peptides and proteins, improving the characterization of the proteins expressed in the microbial communities.

This paper presents a framework for modelling biological tissues based on discrete particles. Cell components (e.g. cell membranes, cell cytoskeleton, cell nucleus) and extracellular matrix (e.g. collagen) are represented using collections of particles. Simple particle to particle interaction laws are used to simulate and control complex physical interaction types (e.g. cell-cell adhesion via cadherins, integrin basement membrane attachment, cytoskeletal mechanical properties). Particles may be given the capacity to change their properties and behaviours in response to changes in the cellular microenvironment (e.g., in response to cell-cell signalling or mechanical loadings). Each particle is in effect an ‘agent’, meaning that the agent can sense local environmental information and respond according to pre-determined or stochastic events. The behaviour of the proposed framework is exemplified through several biological problems of ongoing interest. These examples illustrate how the modelling framework allows enormous flexibility for representing the mechanical behaviour of different tissues, and we argue this is a more intuitive approach than perhaps offered by traditional continuum methods. Because of this flexibility, we believe the discrete modelling framework provides an avenue for biologists and bioengineers to explore the behaviour of tissue systems in a computational laboratory., Author Summary Modelling is an important tool in understanding the behaviour of biological tissues. In this paper we advocate a new modelling framework in which cells and tissues are represented by a collection of particles with associated properties. The particles interact with each other and can change their behaviour in response to changes in their environment. We demonstrate how the propose framework can be used to represent the mechanical behaviour of different tissues with much greater flexibility as compared to traditional continuum based methods.

Correlation patterns in multiple sequence alignments of homologous proteins can be exploited to infer information on the three-dimensional structure of their members. The typical pipeline to address this task, which we in this paper refer to as the three dimensions of contact prediction, is to (i) filter and align the raw sequence data representing the evolutionarily related proteins; (ii) choose a predictive model to describe a sequence alignment; (iii) infer the model parameters and interpret them in terms of structural properties, such as an accurate contact map. We show here that all three dimensions are important for overall prediction success. In particular, we show that it is possible to improve significantly along the second dimension by going beyond the pair-wise Potts models from statistical physics, which have hitherto been the focus of the field. These (simple) extensions are motivated by multiple sequence alignments often containing long stretches of gaps which, as a data feature, would be rather untypical for independent samples drawn from a Potts model. Using a large test set of proteins we show that the combined improvements along the three dimensions are as large as any reported to date., Author Summary Proteins are large molecules that living cells make by stringing together building blocks called amino acids or peptides, following their blue-prints in the DNA. Freshly made proteins are typically long, structure-less chains of peptides, but shortly afterwards most of them fold into characteristic structures. Proteins execute many functions in the cell, for which they need to have the right structure, which is therefore very important in determining what the proteins can do. The structure of a protein can be determined by X-ray diffraction and other experimental approaches which are all, to this day, somewhat labor-intensive and difficult. On the other hand, the order of the peptides in a protein can be read off from the DNA blue-print, and such protein sequences are today routinely produced in large numbers. In this paper we show that many similar protein sequences can be used to find information about the structure. The basic approach is to construct a probabilistic model for sequence variability, and then to use the parameters of that model to predict structure in three-dimensional space. The main technical novelty compared to previous contributions in the same general direction is that we use models more directly matched to the data.

Algorithms for comparing protein structure are frequently used for function annotation. By searching for subtle similarities among very different proteins, these algorithms can identify remote homologs with similar biological functions. In contrast, few comparison algorithms focus on specificity annotation, where the identification of subtle differences among very similar proteins can assist in finding small structural variations that create differences in binding specificity. Few specificity annotation methods consider electrostatic fields, which play a critical role in molecular recognition. To fill this gap, this paper describes VASP-E (Volumetric Analysis of Surface Properties with Electrostatics), a novel volumetric comparison tool based on the electrostatic comparison of protein-ligand and protein-protein binding sites. VASP-E exploits the central observation that three dimensional solids can be used to fully represent and compare both electrostatic isopotentials and molecular surfaces. With this integrated representation, VASP-E is able to dissect the electrostatic environments of protein-ligand and protein-protein binding interfaces, identifying individual amino acids that have an electrostatic influence on binding specificity. VASP-E was used to examine a nonredundant subset of the serine and cysteine proteases as well as the barnase-barstar and Rap1a-raf complexes. Based on amino acids established by various experimental studies to have an electrostatic influence on binding specificity, VASP-E identified electrostatically influential amino acids with 100% precision and 83.3% recall. We also show that VASP-E can accurately classify closely related ligand binding cavities into groups with different binding preferences. These results suggest that VASP-E should prove a useful tool for the characterization of specific binding and the engineering of binding preferences in proteins., Author Summary Proteins, the ubiquitous worker molecules of the cell, are a diverse class of molecules that perform very specific tasks. Understanding how proteins achieve specificity is a critical step towards understanding biological systems and a key prerequisite for rationally engineering new proteins. To examine electrostatic influences on specificity in proteins, this paper presents VASP-E, a software tool that generates solid representations of the electrostatic potential fields that surround proteins. VASP-E compares solids with constructive solid geometry, a class of techniques developed first for modeling complex machine parts. We observed that solid representations could quantify the degree of charge complementarity in protein-protein interactions and identify key residues that strengthen or weaken them. VASP-E correctly identified amino acids with established experimental influences on protein-protein binding specificity. We also observed that solid representations of electrostatic fields could identify electrostatic conservations and variations that relate to similarities and differences in binding specificity between proteins and small molecules.

Simian Foamy Virus (SFV) can be transmitted from non-human primates (NHP) to humans. However, there are no documented cases of human to human transmission, and significant differences exist between infection in NHP and human hosts. The mechanism for these between-host differences is not completely understood. In this paper we develop a new Bayesian approach to the detection of APOBEC3-mediated hypermutation, and use it to compare SFV sequences from human and NHP hosts living in close proximity in Bangladesh. We find that human APOBEC3G can induce genetic changes that may prevent SFV replication in infected humans in vivo., Author Summary Simian Foamy Virus (SFV) is a very common retrovirus in monkeys. When an infected monkey bites a human it can transmit the virus to the human; however, there are no documented cases of human to human transmission. There also appear to be significant differences between infection in monkey and human hosts. The reason for these differences in the two hosts is not completely understood. In this paper we show that a family of host defense enzymes called APOBEC3 may prevent replication of SFV in humans. They do this by changing the genome of the virus so that it cannot replicate. Although this same process also happens in monkeys, it appears to happen less than in humans, and the changes that the monkey APOBEC3 enzymes make are less likely to prevent the virus from replicating. We are able to make these inferences by seeing characteristic types of mutations in a collection of virus DNA sequences sampled in Bangladesh. We develop new statistical methodology to do this analysis.

With much trepidation I accepted the great honor and responsibility of becoming Editor-in-Chief of PLOS Computational Biology. I am fully aware of how hard it will be to step into the shoes of Philip Bourne, the Editor-in-Chief of the journal for the last seven years, since it was founded by him, Steven Brenner, and Michael Eisen. We are all deeply appreciative and thankful to Phil for his unique role; and we are grateful that he will be continuing his association with the journal in the future in the role of Founding Editor-in-Chief, aiding and inspiring us and the PLOS Computational Biology community. As a Deputy Editor-in-Chief, I became aware of the true family relationship in the broad PLOS organization and of the devoted and gifted editorial force so nicely fostered by Phil in PLOS Computational Biology. These played a crucial role in helping me decide to accept the invitation to become Editor-in-Chief. I am deeply committed to the underlying principle of free public access to scientific information. In particular, what is truly unique and special about PLOS Computational Biology is that it fulfills this mission while maintaining the highest standards of scientific rigor, originality, and clear biological relevance. As Editor-in-Chief I will do my best to have PLOS Computational Biology continue these traditions. Computational biology is often perceived as a single field; this however is not the case. Like experimental biology, computational biology is enormously broad; the only distinction from experimental biology is the means. This has disadvantages and advantages: the main disadvantage is that conclusions based on computations are often treated by biologists as less conclusive than those based on experiments; the main advantages are that computations allow researchers to analyze vast amounts of data and make testable predications, and they allow researchers to address problems that current experimental methods may not be able to treat. The high quality of papers published in PLOS Computational Biology indicates that the apparent disadvantage is not necessarily there. While they may still need further experimental and computational validation, conclusions based on rigorous computations applied to carefully assembled and curated data, which are backed up by available experimental results, can be as reliable, insightful, and biologically significant as those based on experiments. PLOS Computational Biology is broad; it addresses diverse biological problems. We look forward to further expanding its scope through the inclusion of outstanding methods and resource papers, opening up significant new research directions while retaining and enhancing the strong scientific merit of PLOS Computational Biology publications. We further plan on improving the pace of submissions processing. PLOS Computational Biology is recognized by the community as the premier journal in computational biology. We will strive to continue in this tradition. PLOS Computational Biology is tightly associated with the International Society of Computational Biology (ISCB). We cherish and will continue fostering this association. The Society, its meetings, and the journal all have a common goal: enhancing and promoting excellence in computational biology. Outstanding research with clear biological relevance, which leads to fundamental new insights into important biological problems, is the hallmark of future contributions by our community to biology. As the Editor-in-Chief I shall do my utmost to achieve these goals.

The International Society for Computational Biology (ISCB; http://www.iscb.org) honors a senior scientist each year for his or her outstanding achievements. The ISCB Accomplishment by a Senior Scientist recognizes a leading member of the computational biology community for his or her significant contributions to the field through research, service, and training. The 2013 ISCB Accomplishment by a Senior Scientist Award honors Dr. David Eisenberg of the University of California Los Angeles (UCLA). Dr. Eisenberg (Image 1) was selected the by ISCB's awards committee, which is chaired by Dr. Alfonso Valencia of the Spanish National Cancer Research Center (CNIO) in Madrid. Dr. Eisenberg will receive his awards and deliver a keynote address at the ISCB's 21st annual Intelligent Systems for Molecular Biology (ISMB) meeting. This meeting is being held jointly with the 12th European Conference on Computational Biology and will take place in Berlin, Germany on July 19–23, 2013 (http://www.iscb.org/ismbeccb2013). Image 1. David Eisenberg, UCLA. 2013 ISCB Accomplishment by a Senior Scientist Award: David Eisenberg David Eisenberg's love of medicine and science was cultivated first during his childhood by his father, a gentle and beloved pediatrician. Eisenberg recalled, “Every night after dinner he would make house calls. I saw how appreciated—even loved—he was in our village.” Eisenberg's father also stoked his scientific curiosity by encouraging him to try some experiments in their basement, including attempts to petrify an egg and to grow worms in chocolate. Eisenberg reminisced, “None of these [experiments] worked, but they were fun!” Eisenberg strongly considered following in his father's footsteps and pursuing a career in medicine. With that goal in mind, he focused his undergraduate studies on biochemical sciences at Harvard University. As a sophomore, he was assigned to Dr. John T. Edsall as a tutor. Edsall was a pioneering researcher in the field of biophysical chemistry, and under his guidance, Eisenberg had his first encounter with laboratory research. “In my junior year, he assigned me to read scientific papers, most of which baffled me, and at the end of that year, I started a research project in his lab, which became the subject of my senior thesis,” Eisenberg recounted. “After graduation, Dr. Edsall turned my thesis into a short paper which was published in Science.” In spite of Eisenberg's eye-opening undergraduate research experiences, he applied and was accepted to medical school. Edsall was also trained as a medical doctor, but Eisenberg remembered how “Dr. Edsall convinced me that if my goal was to improve the health of mankind, I might have a greater impact working in biochemistry, than as a practicing physician.” Eisenberg took Edsall's advice to heart and “finessed making an immediate choice by going to Oxford to study theoretical chemistry under Dr. Charles Coulson, one of the founders of quantum chemistry.” Edsall's guidance had also given him a strong foundation in math and physics, which served him well as a graduate student at Oxford as he recalled being “(just) able to work with Coulson on the energetics of hydrogen bonding.” Eisenberg's postdoctoral studies took him to Princeton in 1964 to work with Dr. Walter Kauzmann, well known for his discovery of the hydrophobic interaction. Eisenberg recollected his ambitious postdoctoral plan “to compute the energy of the hydrophobic interaction in myoglobin, the first protein with a known 3D structure. This plan now seems hopelessly naive: computers were not yet up to such a calculation, potential functions and theory had not advanced to the point that this was a practical problem, and the early protein crystallographers were not eager to release their atomic coordinates.” In light of these challenges, Eisenberg's work with Kauzmann culminated in “a monograph on ice and water, which, incidentally, is still in print 44 years later.” His failed postdoctoral research plan also opened his eyes. He knew that if he wanted to pursue protein energetics, which required knowing protein coordinates, he had to learn X-ray crystallography. Eisenberg's next postdoc took him “to Caltech to study X-ray crystallography with Richard Dickerson, who had been part of the team who had determined the structure of myoglobin.” His X-ray crystallography training was pivotal to establishing his own lab at UCLA, which focused on studying diverse protein structures. Melittin, a component of bee venom, was one of the first structures he determined with his then graduate student Tom Terwilliger. Eisenberg vividly recalled that, “At last I was able to get down to energetic calculations on a protein, and came up with the idea of the hydrophobic moment. This and related ideas gave me, for the first time, the feeling that I could make discoveries.” Eisenberg also remembers the excitement of solving the structure of diphtheria toxin dimer, which he worked on with John Collier, Senyon Choe, and Melanie Bennett (Brewer). He recalled the excitement that stemmed from Bennett's observation that “two monomers of the dimer swapped their third domains, and we called this phenomenon “3D domain swapping.” We explored the implications of 3D domain swapping, again calling on my background in energetics. Diphtheria toxin was the first structural example of 3D domain swapping; now there are hundreds.” Eisenberg's work on protein structures awakened his interest in how protein sequences related to 3D structures. While on sabbatical at the Laboratory of Molecular Biology in Cambridge, he worked with Andrew McLachlan and Mike Gribskov to develop methods to examine protein sequences and use profile analysis to predict the presence of potential structural motifs. These studies led to his work on 3D profiles with Jim Bowie and Roland Luethy, which Eisenberg has now seen “applied to many protein problems.” Burkhard Rost, president of the ISCB, considers Eisenberg's work on hydrophobicity profiling as groundbreaking because it “describes an important feature of the constituents of proteins (amino acids), namely their preferences to stay away from the solvent water (hydro = water, phobie = animosity). Many other outstanding, original methods followed for the prediction of protein structure and function; many of those methods were so visionary that they started entire fields of research.” The availability of the first complete genome sequences in the late 1990s inspired Eisenberg's work with “colleague Todd Yeates and our two talented postdoctoral fellows Edward Marcotte and Matteo Pellegrini, [in which] we found we could extract information on protein interactions from sequenced genomes.” These cutting edge studies resulted in several publications that showed how protein function and protein-protein interactions could be predicted from genome sequences. Eisenberg has focused his research over the last decade on studying amyloid-forming proteins. Several neurodegenerative diseases are associated with amyloid-forming proteins, including Alzheimer's, Parkinson's and amyotrophic lateral sclerosis (Lou Gehrig's) disease. “Just before the turn of the century, I realized that amyloid diseases represent the greatest unmet medical problem facing the world,” Eisenberg recounted. “And at the same time, I realized that structural and computational biology, which have illuminated other areas of biomedicine so well, have not been widely applied to the fundamental problems of amyloid disease. In particular, there had been almost no single crystal X-ray studies of amyloid-forming proteins.” The use of computational biology with this structural data has helped support the definition of the “amyloid state” of proteins. “Bioinformatics and computational biology are great partners with structural biology. Using the tools together can be surprisingly powerful,” said Eisenberg. Eisenberg's group has studied the structural basis of how normal proteins convert to amyloid fibrils. They have gained great insight into this conversion process by determining the atomic structures of the spines of many different types of amyloid fibrils. Eisenberg also acknowledges that, “Having several friends afflicted with amyloid disorders is a great inspiration. I would love to be able to help them, and others. If we can, it would validate Dr. Edsall's advice that sometimes biochemists can do as much, or more, to help mankind than physicians.” Eisenberg remains humble about his accomplishments. When asked about being the recipient of the ISCB Senior Scientist Accomplishment Award, he felt “honored, but perhaps over-honored. There are many others who are equally, or more, deserving of this recognition.” But he also recognizes that this award helps highlight the importance of studying amyloid diseases, especially by using the tools of computational biology. Eisenberg speaks warmly of the mentors that have guided and shaped his scientific training. “I was enormously fortunate to find myself in the research groups of four great mentors: John Edsall, Charles Coulson, Walter Kauzmann, and Richard Dickerson, not to mention my father. All were creative scientists, and also humanists. Watching them I saw their pleasure in scientific discovery, and also saw their insistence on fairness to all those involved in the process of science.” Their examples have not only served him well as a scientist, but also as a mentor. Eisenberg delights in working with trainees because he loves “their eagerness to learn and to succeed, and their willingness to think freshly about hard problems.” Eisenberg's scientific curiosity remains insatiable, and when asked for advice to motivate young scientists, his sage answer was “work on fundamental problems, maintain your curiosity, and above all, persevere.”

Cox regression is commonly used to predict the outcome by the time to an event of interest and in addition, identify relevant features for survival analysis in cancer genomics. Due to the high-dimensionality of high-throughput genomic data, existing Cox models trained on any particular dataset usually generalize poorly to other independent datasets. In this paper, we propose a network-based Cox regression model called Net-Cox and applied Net-Cox for a large-scale survival analysis across multiple ovarian cancer datasets. Net-Cox integrates gene network information into the Cox's proportional hazard model to explore the co-expression or functional relation among high-dimensional gene expression features in the gene network. Net-Cox was applied to analyze three independent gene expression datasets including the TCGA ovarian cancer dataset and two other public ovarian cancer datasets. Net-Cox with the network information from gene co-expression or functional relations identified highly consistent signature genes across the three datasets, and because of the better generalization across the datasets, Net-Cox also consistently improved the accuracy of survival prediction over the Cox models regularized by or . This study focused on analyzing the death and recurrence outcomes in the treatment of ovarian carcinoma to identify signature genes that can more reliably predict the events. The signature genes comprise dense protein-protein interaction subnetworks, enriched by extracellular matrix receptors and modulators or by nuclear signaling components downstream of extracellular signal-regulated kinases. In the laboratory validation of the signature genes, a tumor array experiment by protein staining on an independent patient cohort from Mayo Clinic showed that the protein expression of the signature gene FBN1 is a biomarker significantly associated with the early recurrence after 12 months of the treatment in the ovarian cancer patients who are initially sensitive to chemotherapy. Net-Cox toolbox is available at http://compbio.cs.umn.edu/Net-Cox/., Author Summary Network-based computational models are attracting increasing attention in studying cancer genomics because molecular networks provide valuable information on the functional organizations of molecules in cells. Survival analysis mostly with the Cox proportional hazard model is widely used to predict or correlate gene expressions with time to an event of interest (outcome) in cancer genomics. Surprisingly, network-based survival analysis has not received enough attention. In this paper, we studied resistance to chemotherapy in ovarian cancer with a network-based Cox model, called Net-Cox. The experiments confirm that networks representing gene co-expression or functional relations can be used to improve the accuracy and the robustness of survival prediction of outcome in ovarian cancer treatment. The study also revealed subnetwork signatures that are enriched by extracellular matrix receptors and modulators and the downstream nuclear signaling components of extracellular signal-regulators, respectively. In particular, FBN1, which was detected as a signature gene of high confidence by Net-Cox with network information, was validated as a biomarker for predicting early recurrence in platinum-sensitive ovarian cancer patients in laboratory.

The processing characteristics of neurons in the central auditory system are directly shaped by and reflect the statistics of natural acoustic environments, but the principles that govern the relationship between natural sound ensembles and observed responses in neurophysiological studies remain unclear. In particular, accumulating evidence suggests the presence of a code based on sustained neural firing rates, where central auditory neurons exhibit strong, persistent responses to their preferred stimuli. Such a strategy can indicate the presence of ongoing sounds, is involved in parsing complex auditory scenes, and may play a role in matching neural dynamics to varying time scales in acoustic signals. In this paper, we describe a computational framework for exploring the influence of a code based on sustained firing rates on the shape of the spectro-temporal receptive field (STRF), a linear kernel that maps a spectro-temporal acoustic stimulus to the instantaneous firing rate of a central auditory neuron. We demonstrate the emergence of richly structured STRFs that capture the structure of natural sounds over a wide range of timescales, and show how the emergent ensembles resemble those commonly reported in physiological studies. Furthermore, we compare ensembles that optimize a sustained firing code with one that optimizes a sparse code, another widely considered coding strategy, and suggest how the resulting population responses are not mutually exclusive. Finally, we demonstrate how the emergent ensembles contour the high-energy spectro-temporal modulations of natural sounds, forming a discriminative representation that captures the full range of modulation statistics that characterize natural sound ensembles. These findings have direct implications for our understanding of how sensory systems encode the informative components of natural stimuli and potentially facilitate multi-sensory integration., Author Summary We explore a fundamental question with regard to the representation of sound in the auditory system, namely: what are the coding strategies that underlie observed neurophysiological responses in central auditory areas? There has been debate in recent years as to whether neural ensembles explicitly minimize their propensity to fire (the so-called sparse coding hypothesis) or whether neurons exhibit strong, sustained firing rates when processing their preferred stimuli. Using computational modeling, we directly confront issues raised in this debate, and our results suggest that not only does a sustained firing strategy yield a sparse representation of sound, but the principle yields emergent neural ensembles that capture the rich structural variations present in natural stimuli. In particular, spectro-temporal receptive fields (STRFs) have been widely used to characterize the processing mechanisms of central auditory neurons and have revealed much about the nature of sound processing in central auditory areas. In our paper, we demonstrate how neurons that maximize a sustained firing objective yield STRFs akin to those commonly measured in physiological studies, capturing a wide range of aspects of natural sounds over a variety of timescales, suggesting that such a coding strategy underlies observed neural responses.

An integrative cell migration model incorporating focal adhesion (FA) dynamics, cytoskeleton and nucleus remodeling, actin motor activity, and lamellipodia protrusion is developed for predicting cell spreading and migration behaviors. This work is motivated by two experimental works: (1) cell migration on 2-D substrates under various fibronectin concentrations and (2) cell spreading on 2-D micropatterned geometries. These works suggest (1) cell migration speed takes a maximum at a particular ligand density (∼1140 molecules/µm2) and (2) that strong traction forces at the corners of the patterns may exist due to combined effects exerted by actin stress fibers (SFs). The integrative model of this paper successfully reproduced these experimental results and indicates the mechanism of cell migration and spreading. In this paper, the mechanical structure of the cell is modeled as having two elastic membranes: an outer cell membrane and an inner nuclear membrane. The two elastic membranes are connected by SFs, which are extended from focal adhesions on the cortical surface to the nuclear membrane. In addition, the model also includes ventral SFs bridging two focal adhesions on the cell surface. The cell deforms and gains traction as transmembrane integrins distributed over the outer cell membrane bond to ligands on the ECM surface, activate SFs, and form focal adhesions. The relationship between the cell migration speed and fibronectin concentration agrees with existing experimental data for Chinese hamster ovary (CHO) cell migrations on fibronectin coated surfaces. In addition, the integrated model is validated by showing persistent high stress concentrations at sharp geometrically patterned edges. This model will be used as a predictive model to assist in design and data processing of upcoming microfluidic cell migration assays., Author Summary Cell migration is a complex, multifaceted process, triggered by chemotaxis and haptotatic responses from the extracellular matrix (ECM). It is triggered by a thin lamellipodium protrusion at the leading edge, followed by the assembly of a number of focal adhesions between the lamellipodium base and the ECM. Afterwards, actin stress fibers extend from nascent focal adhesions, some of which connect to the nucleus. In this work, we have developed a dynamic model of cell migration incorporating these four mechanisms of cell biology, such as remodeling of cell and nuclear membranes, focal adhesion dynamics, actin motor activity, and lamellipodia protrusion at the leading edge. We successfully compared our model with existing experimental works of cell migration on (1) substrates with various fibronectin coating concentrations, and (2) cell spreading on three patterned surfaces. Finally, our model demonstrates how actin stress fibers anchored at the trailing edge play a key role, leading to an increase in cell migration speed. Thereby, the model will not only provide new insights on better building such an experiment, but also further experiments will allow us to better validate the model.

Metabolomics is a relatively new “omics” platform, which analyzes a discrete set of metabolites detected in bio-fluids or tissue samples of organisms. It has been used in a diverse array of studies to detect biomarkers and to determine activity rates for pathways based on changes due to disease or drugs. Recent improvements in analytical methodology and large sample throughput allow for creation of large datasets of metabolites that reflect changes in metabolic dynamics due to disease or a perturbation in the metabolic network. However, current methods of comprehensive analyses of large metabolic datasets (metabolomics) are limited, unlike other “omics” approaches where complex techniques for analyzing coexpression/coregulation of multiple variables are applied. This paper discusses the shortcomings of current metabolomics data analysis techniques, and proposes a new multivariate technique (ADEMA) based on mutual information to identify expected metabolite level changes with respect to a specific condition. We show that ADEMA better predicts De Novo Lipogenesis pathway metabolite level changes in samples with Cystic Fibrosis (CF) than prediction based on the significance of individual metabolite level changes. We also applied ADEMA's classification scheme on three different cohorts of CF and wildtype mice. ADEMA was able to predict whether an unknown mouse has a CF or a wildtype genotype with 1.0, 0.84, and 0.9 accuracy for each respective dataset. ADEMA results had up to 31% higher accuracy as compared to other classification algorithms. In conclusion, ADEMA advances the state-of-the-art in metabolomics analysis, by providing accurate and interpretable classification results., Author Summary Metabolomics is an experimental approach that analyzes differences in metabolite levels detected in experimental samples. It has been used in the literature to understand the changes in metabolism with respect to diseases or drugs. Unlike transcriptomics or proteomics, which analyze gene and protein expression levels respectively, the techniques that consider co-regulation of multiple metabolites are quite limited. In this paper, we propose a novel technique, called ADEMA, which computes the expected level changes for each metabolite with respect to a given condition. ADEMA considers multiple metabolites at the same time and is mutual information (MI)-based. We show that ADEMA predicts metabolite level changes for young mice with Cystic Fibrosis (CF) better than significance testing that considers one metabolite at a time. Using three different datasets that contain CF and wild-type (WT) mice, we show that ADEMA can classify an individual as being CF or WT based on the metabolic profiles (with 1.0, 0.84, and 0.9 accuracy, respectively). Compared to other well-known classification algorithms, ADEMA's accuracy is higher by up to 31%.

This paper shows that the various computations underlying spatial cognition can be implemented using statistical inference in a single probabilistic model. Inference is implemented using a common set of ‘lower-level’ computations involving forward and backward inference over time. For example, to estimate where you are in a known environment, forward inference is used to optimally combine location estimates from path integration with those from sensory input. To decide which way to turn to reach a goal, forward inference is used to compute the likelihood of reaching that goal under each option. To work out which environment you are in, forward inference is used to compute the likelihood of sensory observations under the different hypotheses. For reaching sensory goals that require a chaining together of decisions, forward inference can be used to compute a state trajectory that will lead to that goal, and backward inference to refine the route and estimate control signals that produce the required trajectory. We propose that these computations are reflected in recent findings of pattern replay in the mammalian brain. Specifically, that theta sequences reflect decision making, theta flickering reflects model selection, and remote replay reflects route and motor planning. We also propose a mapping of the above computational processes onto lateral and medial entorhinal cortex and hippocampus., Author Summary The ability of mammals to navigate is well studied, both behaviourally and in terms on the underlying neurophysiology. Navigation is a well studied topic in computational fields such as machine learning and signal processing. However, studies in computational neuroscience, which draw together these findings, have mainly focused on specific navigation tasks such as spatial localisation. In this paper, we propose a single probabilistic model which can support multiple tasks, from working out which environment you are in, to computing a sequence of motor commands that will take you to a sensory goal, such as being warm or viewing a particular object. We describe how these tasks can be implemented using a common set of lower level algorithms that implement ‘forward and backward inference over time’. We relate these algorithms to recent findings in animal electrophysiology, where sequences of hippocampal cell activations are observed before, during or after a navigation task, and these sequences are played either forwards or backwards. Additionally, one function of the hippocampus that is preserved across mammals is that it integrates spatial and non-spatial information, and we propose how the forward and backward inference algorithms naturally map onto this architecture.

The protein kinases are a large family of enzymes that play fundamental roles in propagating signals within the cell. Because of the high degree of binding site similarity shared among protein kinases, designing drug compounds with high specificity among the kinases has proven difficult. However, computational approaches to comparing the 3-dimensional geometry and physicochemical properties of key binding site residue positions have been shown to be informative of inhibitor selectivity. The Combinatorial Clustering Of Residue Position Subsets (ccorps) method, introduced here, provides a semi-supervised learning approach for identifying structural features that are correlated with a given set of annotation labels. Here, ccorps is applied to the problem of identifying structural features of the kinase atp binding site that are informative of inhibitor binding. ccorps is demonstrated to make perfect or near-perfect predictions for the binding affinity profile of 8 of the 38 kinase inhibitors studied, while only having overall poor predictive ability for 1 of the 38 compounds. Additionally, ccorps is shown to identify shared structural features across phylogenetically diverse groups of kinases that are correlated with binding affinity for particular inhibitors; such instances of structural similarity among phylogenetically diverse kinases are also shown to not be rare among kinases. Finally, these function-specific structural features may serve as potential starting points for the development of highly specific kinase inhibitors., Author Summary The kinases are a group of essential signaling proteins within the cell and are the largest family of enzymes encoded by the human genome. The high degree of binding site similarity shared across the protein kinases has made them difficult targets for which to design highly selective inhibitors, but kinome-wide binding site analysis can help predict unintended off-target inhibitions. Given the increasingly large number of available kinase structures, kinome-wide comparative analysis of binding sites is now possible. In this paper, the Combinatorial Clustering Of Residue Position Subsets (ccorps) method is introduced and used to synthesize kinome-wide structure datasets with a kinome-wide inhibitor affinity screening dataset consisting of 38 kinase inhibitors. ccorps identifies structural features of the kinase binding site that are correlated with an inhibitor binding and uses these features to predict if this inhibitor will be capable of binding to uncharacterized kinases. This paper demonstrates the ability of ccorps to accurately predict inhibitor binding and identify features of the kinase binding site that are unique to kinases capable of binding a given inhibitor.

The free-energy principle has recently been proposed as a unified Bayesian account of perception, learning and action. Despite the inextricable link between emotion and cognition, emotion has not yet been formulated under this framework. A core concept that permeates many perspectives on emotion is valence, which broadly refers to the positive and negative character of emotion or some of its aspects. In the present paper, we propose a definition of emotional valence in terms of the negative rate of change of free-energy over time. If the second time-derivative of free-energy is taken into account, the dynamics of basic forms of emotion such as happiness, unhappiness, hope, fear, disappointment and relief can be explained. In this formulation, an important function of emotional valence turns out to regulate the learning rate of the causes of sensory inputs. When sensations increasingly violate the agent's expectations, valence is negative and increases the learning rate. Conversely, when sensations increasingly fulfil the agent's expectations, valence is positive and decreases the learning rate. This dynamic interaction between emotional valence and learning rate highlights the crucial role played by emotions in biological agents' adaptation to unexpected changes in their world., Author Summary Emotion plays a crucial role in the adaptation of humans and other animals to changes in their world. Nevertheless, emotion has been neglected in Bayesian models of learning in non-stationary environments. The free-energy principle has recently been proposed as a unified account of learning, perception and action in biological agents. In this paper, we propose a formal definition of emotional valence (i.e., the positive and negative character of emotion) in terms of the rate of change of free-energy or, under some simplifying assumptions, of prediction error over time. This formalization leads to a straightforward and simple meta-learning scheme that accounts for the complex and reciprocal interaction between cognition and emotion. We instantiate this scheme with an emotional agent who is able to dynamically assign emotional valence to every new state of the world that is visited and to experience basic forms of emotion. Crucially, our hypothetical agent uses emotional valence to dynamically adapt to unexpected changes in the world. The proposed scheme is very general in the sense that it is not tied to any particular generative model of sensory inputs.

Spike-timing-dependent plasticity (STDP) has been observed in many brain areas such as sensory cortices, where it is hypothesized to structure synaptic connections between neurons. Previous studies have demonstrated how STDP can capture spiking information at short timescales using specific input configurations, such as coincident spiking, spike patterns and oscillatory spike trains. However, the corresponding computation in the case of arbitrary input signals is still unclear. This paper provides an overarching picture of the algorithm inherent to STDP, tying together many previous results for commonly used models of pairwise STDP. For a single neuron with plastic excitatory synapses, we show how STDP performs a spectral analysis on the temporal cross-correlograms between its afferent spike trains. The postsynaptic responses and STDP learning window determine kernel functions that specify how the neuron “sees” the input correlations. We thus denote this unsupervised learning scheme as ‘kernel spectral component analysis’ (kSCA). In particular, the whole input correlation structure must be considered since all plastic synapses compete with each other. We find that kSCA is enhanced when weight-dependent STDP induces gradual synaptic competition. For a spiking neuron with a “linear” response and pairwise STDP alone, we find that kSCA resembles principal component analysis (PCA). However, plain STDP does not isolate correlation sources in general, e.g., when they are mixed among the input spike trains. In other words, it does not perform independent component analysis (ICA). Tuning the neuron to a single correlation source can be achieved when STDP is paired with a homeostatic mechanism that reinforces the competition between synaptic inputs. Our results suggest that neuronal networks equipped with STDP can process signals encoded in the transient spiking activity at the timescales of tens of milliseconds for usual STDP., Author Summary Tuning feature extraction of sensory stimuli is an important function for synaptic plasticity models. A widely studied example is the development of orientation preference in the primary visual cortex, which can emerge using moving bars in the visual field. A crucial point is the decomposition of stimuli into basic information tokens, e.g., selecting individual bars even though they are presented in overlapping pairs (vertical and horizontal). Among classical unsupervised learning models, independent component analysis (ICA) is capable of isolating basic tokens, whereas principal component analysis (PCA) cannot. This paper focuses on spike-timing-dependent plasticity (STDP), whose functional implications for neural information processing have been intensively studied both theoretically and experimentally in the last decade. Following recent studies demonstrating that STDP can perform ICA for specific cases, we show how STDP relates to PCA or ICA, and in particular explains the conditions under which it switches between them. Here information at the neuronal level is assumed to be encoded in temporal cross-correlograms of spike trains. We find that a linear spiking neuron equipped with pairwise STDP requires additional mechanisms, such as a homeostatic regulation of its output firing, in order to separate mixed correlation sources and thus perform ICA.

Stochastic fluctuations in gene expression give rise to cell-to-cell variability in protein levels which can potentially cause variability in cellular phenotype. For TRAIL (TNF-related apoptosis-inducing ligand) variability manifests itself as dramatic differences in the time between ligand exposure and the sudden activation of the effector caspases that kill cells. However, the contribution of individual proteins to phenotypic variability has not been explored in detail. In this paper we use feature-based sensitivity analysis as a means to estimate the impact of variation in key apoptosis regulators on variability in the dynamics of cell death. We use Monte Carlo sampling from measured protein concentration distributions in combination with a previously validated ordinary differential equation model of apoptosis to simulate the dynamics of receptor-mediated apoptosis. We find that variation in the concentrations of some proteins matters much more than variation in others and that precisely which proteins matter depends both on the concentrations of other proteins and on whether correlations in protein levels are taken into account. A prediction from simulation that we confirm experimentally is that variability in fate is sensitive to even small increases in the levels of Bcl-2. We also show that sensitivity to Bcl-2 levels is itself sensitive to the levels of interacting proteins. The contextual dependency is implicit in the mathematical formulation of sensitivity, but our data show that it is also important for biologically relevant parameter values. Our work provides a conceptual and practical means to study and understand the impact of cell-to-cell variability in protein expression levels on cell fate using deterministic models and sampling from parameter distributions., Author Summary Variability among members of a clonal cell population is increasingly recognized as a near-universal characteristic of prokaryotic and eukaryotic cells. Variability can arise from random fluctuations in the biochemical reactions that control gene transcription, protein synthesis or signal transduction networks. For variability in receptor-mediated signaling responses (in the current work, those activated by the death-inducing ligand TRAIL), we can often distinguish between the influence of stochastic processes that occur prior to ligand exposure and those that occur subsequently. One manifestation of prior variability is cell-to-cell differences in protein concentrations, and this paper uses a combination of modeling and experimentation to ask how these differences impact variability in phenotype, specifically with respect to the timing and probability of cell death. We find that fluctuations in multiple proteins contribute jointly to phenotypic variability, that the contributions of specific proteins to phenotypic variability are highly sensitive to the concentrations of other proteins, and that correlations in protein levels (detectable experimentally) also have a measurable impact on phenotype. Our work provides insight into the regulation of apoptosis and also represents a general approach for understanding cell-to-cell variability in signal transduction pathways.

Hepatitis C virus (HCV) p7 is a membrane-associated oligomeric protein harboring ion channel activity. It is essential for effective assembly and release of infectious HCV particles and an attractive target for antiviral intervention. Yet, the self-assembly and molecular mechanism of p7 ion channelling are currently only partially understood. Using molecular dynamics simulations (aggregate time 1.2 µs), we show that p7 can form stable oligomers of four to seven subunits, with a bias towards six or seven subunits, and suggest that p7 self-assembles in a sequential manner, with tetrameric and pentameric complexes forming as intermediate states leading to the final hexameric or heptameric assembly. We describe a model of a hexameric p7 complex, which forms a transiently-open channel capable of conducting ions in simulation. We investigate the ability of the hexameric model to flexibly rearrange to adapt to the local lipid environment, and demonstrate how this model can be reconciled with low-resolution electron microscopy data. In the light of these results, a view of p7 oligomerization is proposed, wherein hexameric and heptameric complexes may coexist, forming minimalist, yet robust functional ion channels. In the absence of a high-resolution p7 structure, the models presented in this paper can prove valuable as a substitute structure in future studies of p7 function, or in the search for p7-inhibiting drugs., Author Summary Hepatitis C remains a serious global health problem affecting more than 2% of the world's population, and current therapies are effective in only a subset of patients, necessitating an ongoing search for new treatments. The p7 viroporin is considered to be an attractive possible drug target, but rational drug design is hampered by the absence of a high-resolution p7 structure. In this paper, we explore possible structures of oligomeric p7 channels, and discuss the strengths and shortcomings of these models with respect to experimentally determined properties, such as pore-lining residues, ion conductance, and compatibility with low-resolution electron microscopy images. Our results present an image of p7 as a rudimentary, minimalistic ion channel, capable of existing in multiple oligomeric states but exhibiting a bias towards hexamers and heptamers. We believe that the work presented here will be valuable for future research by providing plausible 3-dimensional atomic-resolution models for the visualization of the p7 viroporin and serve as a basis for future computational studies.

Rhinolophidae or Horseshoe bats emit long and narrowband calls. Fluttering insect prey generates echoes in which amplitude and frequency shifts are present, i.e. glints. These glints are reliable cues about the presence of prey and also encode certain properties of the prey. In this paper, we propose that these glints, i.e. the dominant glints, are also reliable signals upon which to base prey localization. In contrast to the spectral cues used by many other bats, the localization cues in Rhinolophidae are most likely provided by self-induced amplitude modulations generated by pinnae movement. Amplitude variations in the echo not introduced by the moving pinnae can be considered as noise interfering with the localization process. The amplitude of the dominant glints is very stable. Therefore, these parts of the echoes contain very little noise. However, using only the dominant glints potentially comes at a cost. Depending on the flutter rate of the insect, a limited number of dominant glints will be present in each echo giving the bat a limited number of sample points on which to base localization. We evaluate the feasibility of a strategy under which Rhinolophidae use only dominant glints. We use a computational model of the echolocation task faced by Rhinolophidae. Our model includes the spatial filtering of the echoes by the morphology of the sonar apparatus of Rhinolophus rouxii as well as the amplitude modulations introduced by pinnae movements. Using this model, we evaluate whether the dominant glints provide Rhinolophidae with enough information to perform localization. Our simulations show that Rhinolophidae can use dominant glints in the echoes as carriers for self-induced amplitude modulations serving as localization cues. In particular, it is shown that the reduction in noise achieved by using only the dominant glints outweighs the information loss that occurs by sampling the echo., Author Summary Rhinolophidae are echolocating bats that hunt among vegetation. The foliage returns clutter echoes that potentially mask the echoes of insect prey. However, prey introduces frequency and amplitude shifts, called glints, into the echo to which these bats are highly sensitive. Therefore, these glints are used by Rhinolophidae to detect prey and infer its properties. One of the defining characteristic of consecutive dominant glints is that they have a very stable amplitude. This is, consecutive wing beats of the insect produce dominant glints with more of less the same amplitude. Owing to the strategy Rhinolophidae use to locate prey, the stable amplitude of glints makes these parts of the echoes ideal signals to use to locate prey. In this paper, we demonstrate the feasibility of strategy under which Rhinolophidae use only the dominant glints in the echo for locating prey.

Histopathological classification of human tumors relies in part on the degree of differentiation of the tumor sample. To date, there is no objective systematic method to categorize tumor subtypes by maturation. In this paper, we introduce a novel computational algorithm to rank tumor subtypes according to the dissimilarity of their gene expression from that of stem cells and fully differentiated tissue, and thereby construct a phylogenetic tree of cancer. We validate our methodology with expression data of leukemia, breast cancer and liposarcoma subtypes and then apply it to a broader group of sarcomas. This ranking of tumor subtypes resulting from the application of our methodology allows the identification of genes correlated with differentiation and may help to identify novel therapeutic targets. Our algorithm represents the first phylogeny-based tool to analyze the differentiation status of human tumors., Author Summary Gene expression profiling of malignancies is often held to demonstrate genes that are “up-regulated” or “down-regulated”, but the appropriate frame of reference against which observations should be compared has not been determined. Fully differentiated somatic cells arise from stem cells, with changes in gene expression that can be experimentally determined. If cancers arise as the result of an abruption of the differentiation process, then poorly differentiated cancers would have a gene expression more similar to stem cells than to normal differentiated tissue, and well differentiated cancers would have a gene expression more similar to fully differentiated cells than to stem cells. In this paper, we describe a novel computational algorithm that allows orientation of cancer gene expression between the poles of the gene expression of stem cells and of fully differentiated tissue. Our methodology allows the construction of a multi-branched phylogeny of human malignancies and can be used to identify genes related to differentiation as well as novel therapeutic targets.