Methods to detect accelerated evolution have identified many genomic regions with unexpectedly rapid evolution in the human lineage-significantly more than in chimpanzees, our closest living relatives. However, these methods focus on accelerated sequence evolution of short non-coding regions, leaving open the questions of how to identify accelerated evolution of molecular function, as opposed to sequence, and whether accelerated evolution has shaped the human genome more broadly. Here, we introduce a new approach to detect accelerated evolution: Function-Aware Statistical Test for Evolutionary Rates (FASTER). In contrast to previous methods, FASTER can detect not only accelerated evolution of sequence, but also of predicted function, and can be applied to any set of genomic regions. Applying this method to humans and chimpanzees, we identified protein-coding, untranslated (UTR), and non-coding regions with accelerated evolution of function. Across all these genomic levels, we consistently found more acceleration in conserved sites in the human lineage compared to chimpanzee, many of which are predicted to reduce protein stability or chromatin accessibility. Multiple lines of evidence suggest this human-acceleration was driven by positive selection on brain development and cognition which has continued to shape human evolution even in the past several thousand years. Collectively, these results demonstrate the power of genome-wide scans for the evolution of predicted function and specifically suggest that an accelerated rate of reduction in function-including widespread decreases in cis-regulatory activity-may have been a major driver of both ancient and recent human evolution.

Models of conformity and anticonformity have typically focused on cultural traits with unordered variants, such as baby names, strategies (cooperate/defect), or the presence/absence of an innovation. There have been fewer studies of conformity to cultural traits with ordered variants, such as level of cooperation (low, medium, high) or proportion of time spent on a task (0% to 100%). In these studies of ordered cultural traits, conformity is defined as a preference for the mean trait value in a population even if no members of the population have variants near this mean; e.g., 50% of the population has variant 0 and 50% has variant 1, producing a mean of 0.5. Here, we introduce models of conformity to ordered traits, which can be either discrete or continuous. In these models, conformists prefer to adopt more popular cultural variants even if these variants are far from the population mean. To measure a variant's "popularity" in cases where no two individuals share precisely the same variant on a continuum, we introduce a metric called k-dispersal; this takes into account a variant's distance to its k closest neighbors, with more "popular" variants having lower distances to their neighbors. We demonstrate through simulations that conformity to ordered traits need not produce a homogeneous population, as has previously been claimed. Under some combinations of parameter values, conformity sustains substantial trait variation over many generations. Furthermore, anticonformity may produce a high level of polarization.

Transcriptional cis-regulation has emerged as the predominant force underlying the evolution of phenotypic diversity, yet our understanding of it is still rudimentary. While empirical comparative genomic approaches have been quite informative, they also suffer from numerous confounders and limited scalability. Here we propose using machine-learning-based methods that predict cis-regulatory activity from DNA sequence to perform in silico ‘genome transplants’ to predict cis-regulatory features as if the genome from one species had been transplanted into the nuclei of another species. Inference of natural selection from the resulting genome-wide catalogs of cis-regulatory divergence could be far more powerful, efficient, and widely applicable than current empirical approaches, enabling unprecedented insights into the genetic basis of biodiversity across the tree of life.

Detecting positive selection is essential to understanding evolution. Many methods to detect positive selection use simple classifications of genetic variants (e.g. synonymous/nonsynonymous). Here, we propose that these methods can be considered special cases of a more general framework to detect positive selection with any variant effect prediction method. Using evolutionary conservation and deep learning-based variant effect predictions, we apply this framework genome-wide in the human lineage and identify positive selection on protein sequences of genes involved in brain and heart development, chromatin accessibility in binding sites of dozens of transcription factors, and non-coding substitutions in compact genomic regions with reinforcing cell type-specific effects on cis -regulatory activity. Consistently, the dominant theme was positive selection on genes regulating neuronal connectivity, suggesting that fine-scale changes in brain wiring were essential to the evolution of human cognition. Overall, this framework represents a powerful, versatile tool to investigate adaptive evolution across the tree of life. ### Competing Interest Statement The authors have declared no competing interest. National Human Genome Research Institute, https://ror.org/00baak391, R01HG012285 National Heart Lung and Blood Institute, HL171542

The selective allosteric ABL1 inhibitor ABL001 (asciminib) represents a new inhibitory mechanism for BCR–ABL1-driven malignancies, and its efficacy and evolving mechanisms of resistance do not overlap with those of other BCR–ABL1 kinase inhibitors. Current inhibitors targeting the BCR–ABL1 mutation have saved many lives but their application is limited by resistance-driving mutations. Here, the authors report the characterization of ABL001, a new allosteric ABL inhibitor. The compound represents a new inhibitory enzymatic mechanism for BCR–ABL-driven malignancies and could be applied for cases of resistance. The authors note that its efficacy and evolving mechanisms of resistance do not overlap with other BCR–ABL kinase inhibitors. Chronic myeloid leukaemia (CML) is driven by the activity of the BCR–ABL1 fusion oncoprotein. ABL1 kinase inhibitors have improved the clinical outcomes for patients with CML, with over 80% of patients treated with imatinib surviving for more than 10 years1. Second-generation ABL1 kinase inhibitors induce more potent molecular responses in both previously untreated and imatinib-resistant patients with CML2. Studies in patients with chronic-phase CML have shown that around 50% of patients who achieve and maintain undetectable BCR–ABL1 transcript levels for at least 2 years remain disease-free after the withdrawal of treatment3,4. Here we characterize ABL001 (asciminib), a potent and selective allosteric ABL1 inhibitor that is undergoing clinical development testing in patients with CML and Philadelphia chromosome-positive (Ph+) acute lymphoblastic leukaemia. In contrast to catalytic-site ABL1 kinase inhibitors, ABL001 binds to the myristoyl pocket of ABL1 and induces the formation of an inactive kinase conformation. ABL001 and second-generation catalytic inhibitors have similar cellular potencies but distinct patterns of resistance mutations, with genetic barcoding studies revealing pre-existing clonal populations with no shared resistance between ABL001 and the catalytic inhibitor nilotinib. Consistent with this profile, acquired resistance was observed with single-agent therapy in mice; however, the combination of ABL001 and nilotinib led to complete disease control and eradicated CML xenograft tumours without recurrence after the cessation of treatment.

Resistance to cancer therapies presents a significant clinical challenge. Recent studies have revealed intratumoral heterogeneity as a source of therapeutic resistance. However, it is unclear whether resistance is driven predominantly by pre-existing or de novo alterations, in part because of the resolution limits of next-generation sequencing. To address this, we developed a high-complexity barcode library, ClonTracer, which enables the high-resolution tracking of more than 1 million cancer cells under drug treatment. In two clinically relevant models, ClonTracer studies showed that the majority of resistant clones were part of small, pre-existing subpopulations that selectively escaped under therapeutic challenge. Moreover, the ClonTracer approach enabled quantitative assessment of the ability of combination treatments to suppress resistant clones. These findings suggest that resistant clones are present before treatment, which would make up-front therapeutic combinations that target non-overlapping resistance a preferred approach. Thus, ClonTracer barcoding may be a valuable tool for optimizing therapeutic regimens with the goal of curative combination therapies for cancer.

Although single genes underlying several evolutionary adaptations have been identified, the genetic basis of complex, polygenic adaptations has been far more challenging to pinpoint. Here we report that the budding yeast Saccharomyces paradoxus has recently evolved resistance to citrinin, a naturally occurring mycotoxin. Applying a genome-wide test for selection on cis-regulation, we identified five genes involved in the citrinin response that are constitutively up-regulated in S. paradoxus. Four of these genes are necessary for resistance, and are also sufficient to increase the resistance of a sensitive strain when over-expressed. Moreover, cis-regulatory divergence in the promoters of these genes contributes to resistance, while exacting a cost in the absence of citrinin. Our results demonstrate how the subtle effects of individual regulatory elements can be combined, via natural selection, into a complex adaptation. Our approach can be applied to dissect the genetic basis of polygenic adaptations in a wide range of species.

BACKGROUND The JAK-STAT pathway is an important signaling pathway downstream of multiple cytokine and growth factor receptors. Dysregulated JAK-STAT signaling has been implicated in the pathogenesis of multiple human malignancies. OBJECTIVE Given this pivotal role of JAK-STAT dysregulation, it is important to identify patients with an overactive JAK-STAT pathway for possible treatment with JAK inhibitors. METHODS We developed a gene signature assay to detect overactive JAK-STAT signaling. The cancer cell line encyclopedia and associated gene-expression data were used to correlate the activation status of STAT5 with the induction of a set of STAT5 target genes. RESULTS Four target genes were identified (PIM1, CISH, SOCS2, and ID1), the expression of which correlated significantly with pSTAT5 status in 40 hematologic tumor cell lines. In pSTAT5-positive models, the expression of the gene signature genes decreased following ruxolitinib treatment, which corresponded to pSTAT5 downmodulation. In pSTAT5-negative cell lines, neither pSTAT5 modulation nor a change in signature gene expression was observed following ruxolitinib treatment. CONCLUSIONS The gene signature can potentially be used to stratify or enrich for patient populations with activated JAK-STAT5 signaling that might benefit from treatments targeting JAK-STAT signaling. Furthermore, the 4-gene signature is a predictor of the pharmacodynamic effects of ruxolitinib.

Background: Chronic myelogenous leukemia (CML) and a subset of acute lymphoblastic leukemia (ALL) are caused by the t(9;22)(q34;q11.2) chromosome translocation, resulting in fusion of the BCR and ABL1 genes on the Philadelphia chromosome to encode constitutively active ABL1 kinase. Despite the dramatic progress made over the past decade with tyrosine kinase inhibitors (TKIs) in the treatment of CML, allogeneic stem cell transplant is considered the only proven curative therapy. To achieve cure or benefit from treatment-free remissions with pharmacologically-based therapies, it is estimated that patients will likely need to achieve a sustained reduction in tumor burden of 4 logs (MR4) or deeper (MR4.5). Currently, only 39% and 18% of patients achieve MR4 by 24 months of treatment with single agent nilotinib or imatinib, respectively. Furthermore, for a subset of CML patients and the majority of Ph+ ALL patients, resistance develops to current TKI’s as a result of emergence of point mutations in the ATP site of the kinase domain. ABL001 is a potent, selective BCR-ABL inhibitor that maintains activity across most mutations, including T315I, with a distinct, allosteric mechanism of action which recently entered Phase I development for the treatment of patients with CML and Ph+ ALL. ABL001 was developed to be dosed in combination with nilotinib to provide greater pharmacological coverage of BCR-ABL disease and prevent the emergence of resistance. Methods: Based on X-ray crystallography, NMR and molecular modeling, ABL001 is the result of a structure-guided medicinal chemistry program targeting the myristoyl pocket of the ABL1 kinase. In vitro cell based assays were performed using the Ba/F3 isogenic cell system and a panel of over 300 cell lines. KCL-22 cells were used to develop an in vivo xenograft model to assess the efficacy of ABL001 and the PD marker, pSTAT5, was used to monitor the inhibition of BCR-ABL signaling. Results: In contrast to TKIs that bind to the ATP-site of the ABL1 kinase domain, NMR and X-Ray crystallography studies confirmed that ABL001 binds to a pocket on the BCR-ABL kinase domain that is normally occupied by the myristoylated N-terminus of ABL1. Upon fusion with BCR, this myristoylated N-terminus that serves to autoregulate ABL1 activity is lost. ABL001 functionally mimics the role of the myristoylated N-terminus by occupying its vacant binding site and restores the negative regulation of the kinase activity. Cell proliferation studies demonstrate that ABL001 selectively inhibited the growth of CML and Ph+ ALL cells with potencies ranging from 1-10nM range. In contrast, BCR-ABL-negative cell lines remained unaffected at concentrations 1000-fold higher. With resistance emerging in the clinic to current TKI’s as a result of point mutations in the ATP-site, ABL001 was tested for activity against clinically observed mutations and found to be active in the low nM range. In the KCL-22 mouse xenograft model, ABL001 displayed potent anti-tumor activity with complete tumor regression observed and a clear dose-dependent correlation with pSTAT5 inhibition. The KCL-22 xenograft model was also used to compare the dosing of ABL001 and nilotinib as single agents to dosing a combination of ABL001 and nilotinib. Single agent dosing regimens led to tumor regressions; however, despite continuous dosing, all tumors relapsed within 30-60 days with evidence of point mutations in the resistant tumors. In contrast, animals treated with the combination of ABL001 and nilotinib achieved sustained tumor regression with no evidence of disease relapse either during the 70 days of treatment or for > 150 days after treatment stopped. Conclusion: ABL001 selectively inhibited the proliferation of cells expressing the BCR-ABL fusion gene and was active against clinically important mutations that arise with current TKI therapy in CML. In an in vivo model of CML, the combination of ABL001 and nilotinib resulted in complete and sustained tumor regression with no evidence of disease relapse. These results provide proof-of-principle that simultaneous targeting of the myristoyl pocket and ATP-pocket by ABL001 and nilotinib, respectively, promotes a more sustained overall efficacy and prevents the emergence of resistance via acquisition of point mutations in the respective binding sites. ABL001 is currently being evaluated in a Phase 1 study in patients with CML and Ph+ ALL. Disclosures Wylie: Novartis Institutes for Biomedical Research, Inc: Employment. Schoepfer:Novartis Institutes for Biomedical Research: Employment. Berellini:Novartis Institutes for Biomedical Research: Employment. Cai:Novartis Institutes for Biomedical Research: Employment. Caravatti:Novartis Institutes for Biomedical Research: Employment. Cotesta:Novartis Institues for Biomedical Research: Employment. Dodd:Novartis Institutes for Biomedical Research: Employment. Donovan:Novartis Institutes for Biomedical Research: Employment. Erb:Novartis Institutes for Biomedical Research: Employment. Furet:Novartis Institutes for Biomedical Research: Employment. Gangal:Novartis Institutes for Biomedical Research: Employment. Grotzfeld:Novartis Institutes for Biomedical Research: Employment. Hassan:Novartis Institutes for Biomedical Research: Employment. Hood:Novartis Institutes for Biomedical Research: Employment. Iyer:Novartis Institutes for Biomedical Research: Employment. Jacob:Novartis Institutes for Biomedical Research: Employment. Jahnke:Novartis Institutes for Biomedical Research: Employment. Lombardo:Novartis Institutes for Biomedical Research: Employment. Loo:Novartis Institutes for Biomedical Research: Employment. Manley:Novartis Institutes for Biomedical Research: Employment. Marzinzik:Novartis Institutes for Biomedical Research: Employment. Palmer:Novartis Institutes for Biomedical Research: Employment. Pelle:Novartis Institutes for Biomedical Research: Employment. Salem:Novartis Institutes for Biomedical Research: Employment. Sharma:Novartis Institutes for Biomedical Research: Employment. Thohan:Novartis Institutes for Biomedical Research: Employment. Zhu:Novartis Institutes for Biomedical Research: Employment. Keen:Novartis Institutes for Biomedical Research: Employment. Petruzzelli:Novartis Institutes for Biomedical Research: Employment. Vanasse:Novartis: Employment, Equity Ownership. Sellers:Novartis: Employment.

It has long been debated whether natural selection acts primarily upon individual organisms, or whether it also commonly acts upon higher-level entities such as lineages. Two arguments against the effectiveness of long-term selection on lineages have been (i) that long-term evolutionary outcomes will not be sufficiently predictable to support a meaningful long-term fitness and (ii) that short-term selection on organisms will almost always overpower long-term selection. Here, we use a computational model of protein folding and binding called ‘lattice proteins’. We quantify the long-term evolutionary success of lineages with two metrics called thek-fitness andk-survivability. We show that long-term outcomes aresurprisingly predictablein this model: only a small fraction of the possible outcomes are ever realized in multiple replicates. Furthermore, the long-term fitness of a lineage depends only partly on its short-term fitness; other factors are also important, including the ‘evolvability’ of a lineage—its capacity to produce adaptive variation. In a system with a distinct short-term and long-term fitness, evolution need not be ‘short-sighted’: lineages may be selected for their long-term properties, sometimes in opposition to short-term selection. Similar evolutionary basins of attraction have been observedin vivo, suggesting that natural biological lineages will also have a predictive long-term fitness.

7111 Background: The JAK/STAT pathway is an important signaling pathway downstream of multiple cytokine and growth factor receptors. Receptor-associated JAKs are activated following receptor-ligand binding. Activated JAKs phosphorylate STAT proteins, which then dimerize and translocate to the nucleus where they modulate the expression of target genes. Dysregulated JAK/STAT signaling has been implicated in the pathogenesis of multiple human malignancies. Activating mutations in JAK2 and the associated activation of STAT5 in myeloproliferative neoplasia is one example of the involvement of this pathway in human cancer. Additionally, overactivated JAK/STAT signaling has been suggested as a survival mechanism in several human cancers. Given the importance of JAK/STAT dysregulation in human diseases, it is important to identify patients with an overactivated JAK/STAT pathway for possible treatment with JAK inhibitors. Thus, we developed a gene signature assay to detect overactivated JAK/STAT5 signaling. Methods: The cancer cell line encyclopedia (CCLE) and associated gene-expression data were used to correlate the activation status of STAT5 with the induction of a set of STAT5 target genes. First, we used 27 tumor cell lines of hematologic lineage, with predetermined phosphorylated STAT5 (pSTAT5) status, to derive STAT5 activation gene signatures. Next, the putative gene signatures were validated against a different set of 13 hematologic tumor cell lines. Results: With this approach, a collection of 7 target genes were identified (PIM1, CISH, SOCS2, ID1, LCN2, EPOR, and EGR1) whose expression significantly correlated with pSTAT5 status in the 40 hematologic tumor cell lines (P < .0001), either together or in specific subsets of 4 and 6 genes (Table). Conclusions: These 4-, 6-, and 7-gene signatures can be used to stratify or select for a patient population with activated JAK/STAT5 signaling that could potentially benefit from treatments targeting the JAK/STAT5 signaling pathway. [Table: see text]

ABSTRACT Simple sequence repeat (SSR) tracts produce stochastic on-off switching, or phase variation, in the expression of a panoply of surface molecules in many bacterial commensals and pathogens. A change to the number of repeats in a tract may alter the phase of the translational reading frame, which toggles the on-off state of the switch. Here, we construct an in silico SSR locus with mutational dynamics calibrated to those of the Haemophilus influenzae mod locus. We simulate its evolution in a regimen of two alternating environments, simultaneously varying the selection coefficient, s, and the epoch length, T. Some recent work in a simpler (two-locus) model suggested that stochastic switching in a regimen of two alternating environments may be evolutionarily favored only if the selection coefficients in the two environments are nearly equal (“symmetric”) or selection is very strong. This finding was puzzling, as it greatly restricted the conditions under which stochastic switching might evolve. Instead, we find agreement with other recent theoretical work, observing selective utility for stochastic switching if the product sT is large enough for the favored state to nearly fix in both environments. Symmetry is required neither in s nor in sT. Because we simulate finite populations and use a detailed model of the SSR locus, we are also able to examine the impact of population size and of several SSR locus parameters. Our results indicate that conditions favoring evolution and maintenance of SSR loci in bacteria are quite broad. IMPORTANCE Bacteria experience frequent changes of environment during the infection cycle. One means to rapidly adapt is stochastic switching: a bacterial lineage will stochastically produce a variety of genotypes, so that some descendants will survive if the environment changes. Stochastic switching mediated by simple sequence repeat (SSR) loci is widespread among bacterial commensals and pathogens and influences critical interactions with host surfaces or immune effectors, thereby affecting host persistence, transmission, and virulence. Here, we use the most detailed in silico model of an SSR locus to date, with its phase variation calibrated to match the mod locus of Haemophilus influenzae. The type III restriction-modification system encoded by mod participates in the regulation of multiple other genes; thus, SSR-mediated phase variation of mod has far-reaching cis-regulatory effects. This coupling of phase-variable switching to complex phenotypic effects has been described as the “phasevarion” and is central to understanding the infection cycle of bacterial commensals and pathogens.

It is my pleasure to welcome you to Amsterdam for the 2013 Genetic and Evolutionary Computation Conference (GECCO-2013). This year GECCO is being held in Europe, and we very much hope you enjoy the city of Amsterdam, as well as the many historical and modern places directly reachable from it. We hope to offer you an appealing combination of research and culture worth experiencing. Organizing GECCO has been for me both a pleasure and an honor. A pleasure, because I have been involved in it since 1999: now a long trip, with this incredible community of skilled people and close friends. An honor, since I am having the possibility to give back to the community what I have got from GECCO, personally and in my own research. The high quality, the ideas, and the possibility to interact with GECCO's attendees had a profound impact in my students, my postdocs, and my actual research program. Christian Blum served as the editor-in-chief this year, and did a superb job in maintaining the well-known high quality of this conference. GECCO-2013 accepted 204 full papers for oral presentation out of a total of 570 submitted. This large number of submissions accounts for the world interest in this year's GECCO. But also, it represents an acceptance rate of less than 36%, what tells on the expected quality of the accepted works. I am very thankful to Christian Blum, Leonardo Vanneschi (this year's Proceedings Chair), and all the track chairs for their hard work managing the review, selection, and scheduling process for the scientific papers. Thanks also to Linklings and Sheridan Printing for their assistance with the publication of the …

Using a model of evolved gene regulatory networks, we illustrate several quantitative metrics relating to the long-term evolution of lineages. The k-generation fitness and k-generation survivability measure the evolutionary success of lineages. An entropy measure is used to quantify the predictability of lineage evolution. The metrics are readily applied to any system in which lineage membership can be periodically counted, and provide a quantitative characterization of the genetic landscape, genotype-phenotype map, and fitness landscape. Evolution is shown to be surprisingly predictable in gene networks: only a small number of the possible outcomes are ever observed in multiple replicate experiments. We emphasize the view that the lineage (not the individual, or the genotype) is the evolving entity over the long term. Notably, the long-term fitness is distinct from the short-term fitness. Since evolution is repeatable over the long-term, this implies long-term selection on lineages is possible; the evolutionary process need not be "short-sighted". If we wish to evolve very complex artifacts, it will be expedient to promote the long-term evolution of the genetic architecture by tailoring our models to emphasize long-term selection.

Using a rule-based system for growing artificial neural networks [1], we have evolved controllers for physically simulated 18-degree-of-freedom robotic "spiders". These robots previously evolved "galloping" gaits and tracked a compass heading. Here, we introduce an artificial "visual cortex" that permits distinct predator and prey species to "see" one another. It has been claimed [2] that the "Cambrian explosion" of evolutionary diversity was due the emergence of animal eyes, which created a new dynamic in the interaction between predators and prey. The survival of both was for the first time determined by interactions between animals some distance apart. Here, we find that the introduction of vision does indeed generate complex interactions both between and within the two species, as they species evolve complex hunting or avoidance behavior, modulating their galloping gaits with dynamic input from their visual systems.

Mutability as mechanism for rapid adaptation to environmental challenge is an alluringly simple concept whose apotheosis is realized in simple sequence repeats (SSR). Bacterial genomes of several species contain SSRs with a proven role in adaptation to environmental fluctuations. SSRs are hypermutable and generate reversible mutations in localized regions of bacterial genomes, leading to phase variable ON/OFF switches in gene expression. The application of genetic, bioinformatic, and mathematical/computational modeling approaches are revolutionizing our current understanding of how genomic molecular forces and environmental factors influence SSR-mediated adaptation and led to evolution of this mechanism of localized hypermutation in bacterial genomes.

For a lineage to survive over long time periods, it must sometimes change. This has given rise to the term evolvability, meaning the tendency to produce adaptive variation. One lineage may be superior to another in terms of its current standing variation, or it may tend to produce more adaptive variation. However, evolutionary outcomes depend on more than standing variation and produced adaptive variation: deleterious variation also matters. Evolvability, as most commonly interpreted, is not predictive of evolutionary outcomes. Here, we define a predictive measure of the evolutionary success of a lineage that we call the k-survivability, defined as the probability that the lineage avoids extinction for k generations. We estimate the k-survivability using multiple experimental replicates. Because we measure evolutionary outcomes, the initial standing variation, the full spectrum of generated variation, and the heritability of that variation are all incorporated. Survivability also accounts for the decreased joint likelihood of extinction of sub-lineages when they 1) disperse in space, or 2) diversify in lifestyle. We illustrate measurement of survivability with in silico models, and suggest that it may also be measured in vivo using multiple longitudinal replicates. The k-survivability is a metric that enables the quantitative study of, for example, the evolution of 1) mutation rates, 2) dispersal mechanisms, 3) the genotype-phenotype map, and 4) sexual reproduction, in temporally and spatially fluctuating environments. Although these disparate phenomena evolve by well-understood microevolutionary rules, they are also subject to the macroevolutionary constraint of long-term survivability.

Using a rule-based system for growing artificial neural networks, we have evolved controllers for physically simulated robotic "spiders". The controllers take their input from an “artificial retina” that senses other spiders and inanimate barrier objects in the environment, and must provide output to dynamically control the 18 degrees of freedom of the six legs of the robot every time step. We perform evolutionary runs with two species of spider that interact in simulation with each other and with inanimate barrier objects. One species (the "predator") is selectively rewarded for "eating" (by physically colliding with) the other species, and the other (the "prey") is selectively penalized for being caught, and rewarded for "eating" the barriers. The two species evolve complex running gaits, with control inputs coming from their retinas that produce hunting or avoidance behavior. We suggest that predator-prey frequency dependent selection can provide a relatively long-term genetic memory of previously searched regions of phenotype space, enforcing a form of novelty search that may reduce duplicated evolutionary search effort.