Vocal Learning in Songbirds and Humans: A Retrospective in Honor of Peter Marler

Peter Marler made a number of significant contributions to the field of ethology, particularly in the area of animal communication. His research on birdsong learning gave rise to a thriving subfield. An important tenet of this growing subfield is that parallels between birdsong and human speech make songbirds valuable as models in comparative and translational research, particularly in the case of vocal learning and development. Decades ago, Marler pointed out several phenomena common to the processes of vocal development in songbirds and humans—including a dependence on early acoustic experience, sensitive periods, predispositions, auditory feedback, intrinsic reinforcement, and a progression through distinct developmental stages—and he advocated for the value of comparative study in this domain. We review Marler's original comparisons between birdsong and speech ontogeny and summarize subsequent progress in research into these and other parallels. We also revisit Marler's arguments in support of the comparative study of vocal development in the context of its widely recognized value today.     

Avian vocal mimicry: a unified conceptual framework

Mimicry is a classical example of adaptive signal design. Here, we review the current state of research into vocal mimicry in birds. Avian vocal mimicry is a conspicuous and often spectacular form of animal communication, occurring in many distantly related species. However, the proximate and ultimate causes of vocal mimicry are poorly understood. In the first part of this review, we argue that progress has been impeded by conceptual confusion over what constitutes vocal mimicry. We propose a modified version of Vane‐Wright's (1980) widely used definition of mimicry. According to our definition, a vocalisation is mimetic if the behaviour of the receiver changes after perceiving the acoustic resemblance between the mimic and the model, and the behavioural change confers a selective advantage on the mimic. Mimicry is therefore specifically a functional concept where the resemblance between heterospecific sounds is a target of selection. It is distinct from other forms of vocal resemblance including those that are the result of chance or common ancestry, and those that have emerged as a by‐product of other processes such as ecological convergence and selection for large song‐type repertoires. Thus, our definition provides a general and functionally coherent framework for determining what constitutes vocal mimicry, and takes account of the diversity of vocalisations that incorporate heterospecific sounds. In the second part we assess and revise hypotheses for the evolution of avian vocal mimicry in the light of our new definition. Most of the current evidence is anecdotal, but the diverse contexts and acoustic structures of putative vocal mimicry suggest that mimicry has multiple functions across and within species. There is strong experimental evidence that vocal mimicry can be deceptive, and can facilitate parasitic interactions. There is also increasing support for the use of vocal mimicry in predator defence, although the mechanisms are unclear. Less progress has been made in explaining why many birds incorporate heterospecific sounds into their sexual displays, and in determining whether these vocalisations are functionally mimetic or by‐products of sexual selection for other traits such as repertoire size. Overall, this discussion reveals a more central role for vocal mimicry in the behavioural ecology of birds than has previously been appreciated. The final part of this review identifies important areas for future research. Detailed empirical data are needed on individual species, including on the structure of mimetic signals, the contexts in which mimicry is produced, how mimicry is acquired, and the ecological relationships between mimic, model and receiver. At present, there is little information and no consensus about the various costs of vocal mimicry for the protagonists in the mimicry complex. The diversity and complexity of vocal mimicry in birds raises important questions for the study of animal communication and challenges our view of the nature of mimicry itself. Therefore, a better understanding of avian vocal mimicry is essential if we are to account fully for the diversity of animal signals.     

Developmental shifts in gene expression in the auditory forebrain during the sensitive period for song learning

A male zebra finch begins to learn to sing by memorizing a tutor's song during a sensitive period in juvenile development. Tutor song memorization requires molecular signaling within the auditory forebrain. Using microarray and in situ hybridizations, we tested whether the auditory forebrain at an age just before tutoring expresses a different set of genes compared with later life after song learning has ceased. Microarray analysis revealed differences in expression of thousands of genes in the male auditory forebrain at posthatch day 20 (P20) compared with adulthood. Furthermore, song playbacks had essentially no impact on gene expression in P20 auditory forebrain, but altered expression of hundreds of genes in adults. Most genes that were song‐responsive in adults were expressed at constitutively high levels at P20. Using in situ hybridization with a representative sample of 44 probes, we confirmed these effects and found that birds at P20 and P45 were similar in their gene expression patterns. Additionally, eight of the probes showed male–female differences in expression. We conclude that the developing auditory forebrain is in a very different molecular state from the adult, despite its relatively mature gross morphology and electrophysiological responsiveness to song stimuli. Developmental gene expression changes may contribute to fine‐tuning of cellular and molecular properties necessary for song learning. © 2009 Wiley Periodicals, Inc. Develop Neurobiol 2009     

The learning advantage: bird species that learn their song show a tighter adjustment of song to noisy environments than those that do not learn

Song learning has evolved within several avian groups. Although its evolutionary advantage is not clear, it has been proposed that song learning may be advantageous in allowing birds to adapt their songs to the local acoustic environment. To test this hypothesis, we analysed patterns of song adjustment to noisy environments and explored their possible link to song learning. Bird vocalizations can be masked by low‐frequency noise, and birds respond to this by singing higher‐pitched songs. Most reports of this strategy involve oscines, a group of birds with learning‐based song variability, and it is doubtful whether species that lack song learning (e.g. suboscines) can adjust their songs to noisy environments. We address this question by comparing the degree of song adjustment to noise in a large sample of oscines (17 populations, 14 species) and suboscines (11 populations, 7 species), recorded in Brazil (Manaus, Brasilia and Curitiba) and Mexico City. We found a significantly stronger association between minimum song frequency and noise levels (effect size) in oscines than in suboscines, suggesting a tighter match in oscines between song transmission capacity and ambient acoustics. Suboscines may be more vulnerable to acoustic pollution than oscines and thus less capable of colonizing cities or acoustically novel habitats. Additionally, we found that species whose song frequency was more divergent between populations showed tighter noise–song frequency associations. Our results suggest that song learning and/or song plasticity allows adaptation to new habitats and that this selective advantage may be linked to the evolution of song learning and plasticity.     

Memory in the making: localized brain activation related to song learning in young songbirds

Songbird males learn to sing their songs from an adult ‘tutor’ early in life, much like human infants learn to speak. Similar to humans, in the songbird brain there are separate neural substrates for vocal production and for auditory memory. In adult songbirds, the caudal pallium, the avian equivalent of the auditory association cortex, has been proposed to contain the neural substrate of tutor song memory, while the song system is involved in song production as well as sensorimotor learning. If this hypothesis is correct, there should be neuronal activation in the caudal pallium, and not in the song system, while the young bird is hearing the tutor song. We found increased song-induced molecular neuronal activation, measured as the expression of an immediate early gene, in the caudal pallium of juvenile zebra finch males that were in the process of learning to sing their songs. No such activation was found in the song system. Molecular neuronal activation was significantly greater in response to tutor song than to novel song or silence in the medial part of the caudomedial nidopallium (NCM). In the caudomedial mesopallium, there was significantly greater molecular neuronal activation in response to tutor song than to silence. In addition, in the NCM there was a significant positive correlation between spontaneous molecular neuronal activation and the strength of song learning during sleep. These results suggest that the caudal pallium contains the neural substrate for tutor song memory, which is activated during sleep when the young bird is in the process of learning its song. The findings provide insight into the formation of auditory memories that guide vocal production learning, a process fundamental for human speech acquisition.     

Perineuronal nets and vocal plasticity in songbirds: A proposed mechanism to explain the difference between closed‐ended and open‐ended learning

Perineuronal nets (PNN) are aggregations of chondroitin sulfate proteoglycans surrounding the soma and proximal processes of neurons, mostly GABAergic interneurons expressing parvalbumin. They limit the plasticity of their afferent synaptic connections. In zebra finches PNN develop in an experience‐dependent manner in the song control nuclei HVC and RA (nucleus robustus arcopallialis) when young birds crystallize their song. Because songbird species that are open‐ended learners tend to recapitulate each year the different phases of song learning until their song crystallizes at the beginning of the breeding season, we tested whether seasonal changes in PNN expression would be found in the song control nuclei of a seasonally breeding species such as the European starling. Only minimal changes in PNN densities and total number of cells surrounded by PNN were detected. However, comparison of the density of PNN and of PNN surrounding parvalbumin‐positive cells revealed that these structures are far less numerous in starlings that show extensive adult vocal plasticity, including learning of new songs throughout the year, than in the closed‐ended learner zebra finches. Canaries that also display some vocal plasticity across season but were never formally shown to learn new songs in adulthood were intermediate in this respect. Together these data suggest that establishment of PNN around parvalbumin‐positive neurons in song control nuclei has diverged during evolution to control the different learning capacities observed in songbird species. This differential expression of PNN in different songbird species could represent a key cellular mechanism mediating species variation between closed‐ended and open‐ended learning strategies. © 2017 Wiley Periodicals, Inc. Develop Neurobiol 77: 975–994, 2017     

From songs to synapses: molecular mechanisms of birdsong memory. Molecular mechanisms of auditory learning in songbirds involve immediate early genes, including zenk and arc, the ERK/MAPK pathway and synapsins

There are remarkable behavioral, neural, and genetic similarities between the way songbirds learn to sing and human infants learn to speak. Furthermore, the brain regions involved in birdsong learning, perception, and production have been identified and characterized in detail. In particular, the caudal medial nidopallium (the avian analog of the mammalian auditory-association cortex) has been found to contain the neural substrate of auditory memory, paving the way for analyses of the underlying molecular mechanisms. Recently, the zebra finch genome was sequenced, and annotated cDNA databases representing over 15,000 unique brain-expressed genes are available, enabling high-throughput gene expression analyses. Here we review the involvement of immediate early genes (e.g. zenk and arc), their downstream targets (e.g. synapsins), and their regulatory signaling pathways (e.g. MAPK/ERK) in songbird memory. We propose that in-depth investigations of zenk- and ERK-dependent cascades will help to further unravel the molecular basis of auditory memory.     

Automated analysis of behavior: A computer‐controlled system for drug screening and the investigation of learning

Efforts to understand cognition will be greatly facilitated by computerized systems that enable the automated analysis of animal behavior. A number of controversies in the invertebrate learning field have resulted from difficulties inherent in manual experiments. Driven by the necessity to overcome these problems during investigation of neural function in planarian flatworms and frog larvae, we designed and developed a prototype for an inexpensive, flexible system that enables automated control and analysis of behavior and learning. Applicable to a variety of small animals such as flatworms and zebrafish, this system allows automated analysis of innate behavior, as well as of learning and memory in a plethora of conditioning paradigms. We present here the schematics of a basic prototype, which overcomes experimenter effects and operator tedium, enabling a large number of animals to be analyzed with transparent on‐line access to primary data. A scaled‐up version of this technology represents an efficient methodology to screen pharmacological and genetic libraries for novel neuroactive reagents of basic and biomedical relevance. © 2006 Wiley Periodicals, Inc. J Neurobiol, 2006     

Deep learning methods to generate synthetic CT from MRI in radiotherapy: A literature review

PurposeIn radiotherapy, MRI is used for target volume and organs-at-risk delineation for its superior soft-tissue contrast as compared to CT imaging. However, MRI does not provide the electron density of tissue necessary for dose calculation. Several methods of synthetic-CT (sCT) generation from MRI data have been developed for radiotherapy dose calculation. This work reviewed deep learning (DL) sCT generation methods and their associated image and dose evaluation, in the context of MRI-based dose calculation.MethodsWe searched the PubMed and ScienceDirect electronic databases from January 2010 to March 2021. For each paper, several items were screened and compiled in figures and tables.ResultsThis review included 57 studies. The DL methods were either generator-only based (45% of the reviewed studies), or generative adversarial network (GAN) architecture and its variants (55% of the reviewed studies). The brain and pelvis were the most commonly investigated anatomical localizations (39% and 28% of the reviewed studies, respectively), and more rarely, the head-and-neck (H&N) (15%), abdomen (10%), liver (5%) or breast (3%). All the studies performed an image evaluation of sCTs with a diversity of metrics, with only 36 studies performing dosimetric evaluations of sCT.ConclusionsThe median mean absolute errors were around 76 HU for the brain and H&N sCTs and 40 HU for the pelvis sCTs. For the brain, the mean dose difference between the sCT and the reference CT was <2%. For the H&N and pelvis, the mean dose difference was below 1% in most of the studies. Recent GAN architectures have advantages compared to generator-only, but no superiority was found in term of image or dose sCT uncertainties. Key challenges of DL-based sCT generation methods from MRI in radiotherapy is the management of movement for abdominal and thoracic localizations, the standardization of sCT evaluation, and the investigation of multicenter impacts.     

O‐GlcNAcylation of protein kinase A catalytic subunits enhances its activity: a mechanism linked to learning and memory deficits in Alzheimer's disease

Alzheimer's disease (AD) is characterized clinically by memory loss and cognitive decline. Protein kinase A (PKA)‐CREB signaling plays a critical role in learning and memory. It is known that glucose uptake and O‐GlcNAcylation are reduced in AD brain. In this study, we found that PKA catalytic subunits (PKAcs) were posttranslationally modified by O‐linked N‐acetylglucosamine (O‐GlcNAc). O‐GlcNAcylation regulated the subcellular location of PKAcα and PKAcβ and enhanced their kinase activity. Upregulation of O‐GlcNAcylation in metabolically active rat brain slices by O‐(2‐acetamido‐2‐deoxy‐d ‐glucopyranosylidenamino) N‐phenylcarbamate (PUGNAc), an inhibitor of N‐acetylglucosaminidase, increased the phosphorylation of tau at the PKA site, Ser214, but not at the non‐PKA site, Thr205. In contrast, in rat and mouse brains, downregulation of O‐GlcNAcylation caused decreases in the phosphorylation of CREB at Ser133 and of tau at Ser214, but not at Thr205. Reduction in O‐GlcNAcylation through intracerebroventricular injection of 6‐diazo‐5‐oxo‐l ‐norleucine (DON), the inhibitor of glutamine fructose‐6‐phosphate amidotransferase, suppressed PKA‐CREB signaling and impaired learning and memory in mice. These results indicate that in addition to cAMP and phosphorylation, O‐GlcNAcylation is a novel mechanism that regulates PKA‐CREB signaling. Downregulation of O‐GlcNAcylation suppresses PKA‐CREB signaling and consequently causes learning and memory deficits in AD.     

A deep learning approach for converting prompt gamma images to proton dose distributions: A Monte Carlo simulation study

PurposeIn proton therapy, imaging prompt gamma (PG) rays has the potential to verify proton dose (PD) distribution. Despite the fact that there is a strong correlation between the gamma-ray emission and PD, they are still different in terms of the distribution and the Bragg peak (BP) position. In this work, we investigated the feasibility of using a deep learning approach to convert PG images to PD distributions.MethodsWe designed the Monte Carlo simulations using 20 digital brain phantoms irradiated with a 100-MeV proton pencil beam. Each phantom was used to simulate 200 pairs of PG images and PD distributions. A convolutional neural network based on the U-net architecture was trained to predict PD distributions from PG images.ResultsOur simulation results show that the pseudo PD distributions derived from the corresponding PG images agree well with the simulated ground truths. The mean of the BP position errors from each phantom was less than 0.4 mm. We also found that 2000 pairs of PG images and dose distributions would be sufficient to train the U-net. Moreover, the trained network could be deployed on the unseen data (i.e. different beam sizes, proton energies and real patient CT data).ConclusionsOur simulation study has shown the feasibility of predicting PD distributions from PG images using a deep learning approach, but the reliable prediction of PD distributions requires high-quality PG images. Image-degrading factors such as low counts and limited spatial resolution need to be considered in order to obtain high-quality PG images.     

CLADES: A classification‐based machine learning method for species delimitation from population genetic data

Species are considered to be the basic unit of ecological and evolutionary studies. As multilocus genomic data are increasingly available, there have been considerable interests in the use of DNA sequence data to delimit species. In this study, we show that machine learning can be used for species delimitation. Our method treats the species delimitation problem as a classification problem for identifying the category of a new observation on the basis of training data. Extensive simulation is first conducted over a broad range of evolutionary parameters for training purposes. Each pair of known populations is combined to form training samples with a label of “same species” or “different species”. We use support vector machine (SVM) to train a classifier using a set of summary statistics computed from training samples as features. The trained classifier can classify a test sample to two outcomes: “same species” or “different species”. Given multilocus genomic data of multiple related organisms or populations, our method (called CLADES) performs species delimitation by first classifying pairs of populations. CLADES then delimits species by maximizing the likelihood of species assignment for multiple populations. CLADES is evaluated through extensive simulation and also tested on real genetic data. We show that CLADES is both accurate and efficient for species delimitation when compared with existing methods. CLADES can be useful especially when existing methods have difficulty in delimitation, for example with short species divergence time and gene flow.     

A case for project based learning to enact semantic waves: towards cumulative knowledge building

ABSTRACT

First-year undergraduate curricula and their delivery should assist students in the transition from previous learning experiences to learning in higher education. However, the so-called articulation gap or discontinuity between secondary and higher education has been identified as a key structural curriculum problem for first-year success in South Africa and abroad. Valuable insights into this problem came from a recent study that drew on Legitimation Code Theory (LCT). Findings revealed an unexpectedly wide gap between the high school and the university biology curricula. The high school biology curriculum displays minimal movement between context-dependent, simpler meaning and relatively decontextualized, condensed meaning common in first-year biology. LCT Semantics was also found to be a valuable tool for restructuring curricula and pedagogy to intentionally enact semantic movement and thereby a more gradual transition for students from high school to university. This paper reports on an integrative first-year biology project aimed intentionally at taking students’ concept knowledge through a wide contextual range, and repeatedly between less and more complex meaning. I reflect on how the project design steers students towards creating semantic movement during their presentations, thereby contributing to cumulative knowledge building and a more gradual transition towards first-year epistemological access.     

EDock‐ML: A web server for using ensemble docking with machine learning to aid drug discovery

EDock‐ML is a web server that facilitates the use of ensemble docking with machine learning to help decide whether a compound is worthwhile to be considered further in a drug discovery process. Ensemble docking provides an economical way to account for receptor flexibility in molecular docking. Machine learning improves the use of the resulting docking scores to evaluate whether a compound is likely to be useful. EDock‐ML takes a bottom‐up approach in which machine‐learning models are developed one protein at a time to improve predictions for the proteins included in its database. Because the machine‐learning models are intended to be used without changing the docking and model parameters with which the models were trained, novice users can use it directly without worrying about what parameters to choose. A user simply submits a compound specified by an ID from the ZINC database (Sterling, T.; Irwin, J. J., J Chem Inf Model 2015, 55[11], 2,324–2,337.) or upload a file prepared by a chemical drawing program and receives an output helping the user decide the likelihood of the compound to be active or inactive for a drug target. EDock‐ML can be accessed freely at edock‐ml.umsl.edu     

A deep learning approach to conflating heterogeneous geospatial data for corn yield estimation: A case study of the US Corn Belt at the county level

Understanding large‐scale crop growth and its responses to climate change are critical for yield estimation and prediction, especially under the increased frequency of extreme climate and weather events. County‐level corn phenology varies spatially and interannually across the Corn Belt in the United States, where precipitation and heat stress presents a temporal pattern among growth phases (GPs) and vary interannually. In this study, we developed a long short‐term memory (LSTM) model that integrates heterogeneous crop phenology, meteorology, and remote sensing data to estimate county‐level corn yields. By conflating heterogeneous phenology‐based remote sensing and meteorological indices, the LSTM model accounted for 76% of yield variations across the Corn Belt, improved from 39% of yield variations explained by phenology‐based meteorological indices alone. The LSTM model outperformed least absolute shrinkage and selection operator (LASSO) regression and random forest (RF) approaches for end‐of‐the‐season yield estimation, as a result of its recurrent neural network structure that can incorporate cumulative and nonlinear relationships between corn yield and environmental factors. The results showed that the period from silking to dough was most critical for crop yield estimation. The LSTM model presented a robust yield estimation under extreme weather events in 2012, which reduced the root‐mean‐square error to 1.47 Mg/ha from 1.93 Mg/ha for LASSO and 2.43 Mg/ha for RF. The LSTM model has the capability to learn general patterns from high‐dimensional (spectral, spatial, and temporal) input features to achieve a robust county‐level crop yield estimation. This deep learning approach holds great promise for better understanding the global condition of crop growth based on publicly available remote sensing and meteorological data.     

A Proboscis Extension Response Protocol for Investigating Behavioral Plasticity in Insects: Application to Basic,Biomedical, and Agricultural Research

Insects modify their responses to stimuli through experience of associating those stimuli with events important for survival (e.g., food, mates, threats). There are several behavioral mechanisms through which an insect learns salient associations and relates them to these events. It is important to understand this behavioral plasticity for programs aimed toward assisting insects that are beneficial for agriculture. This understanding can also be used for discovering solutions to biomedical and agricultural problems created by insects that act as disease vectors and pests. The Proboscis Extension Response (PER) conditioning protocol was developed for honey bees (Apis mellifera) over 50 years ago to study how they perceive and learn about floral odors, which signal the nectar and pollen resources a colony needs for survival. The PER procedure provides a robust and easy-to-employ framework for studying several different ecologically relevant mechanisms of behavioral plasticity. It is easily adaptable for use with several other insect species and other behavioral reflexes. These protocols can be readily employed in conjunction with various means for monitoring neural activity in the CNS via electrophysiology or bioimaging, or for manipulating targeted neuromodulatory pathways. It is a robust assay for rapidly detecting sub-lethal effects on behavior caused by environmental stressors, toxins or pesticides.We show how the PER protocol is straightforward to implement using two procedures. One is suitable as a laboratory exercise for students or for quick assays of the effect of an experimental treatment. The other provides more thorough control of variables, which is important for studies of behavioral conditioning. We show how several measures for the behavioral response ranging from binary yes/no to more continuous variable like latency and duration of proboscis extension can be used to test hypotheses. And, we discuss some pitfalls that researchers commonly encounter when they use the procedure for the first time.     

A machine learning approach for the identification of population-informative markers from high-throughput genotyping data: application to several pig breeds

Single nucleotide polymorphisms (SNPs) able to describe population differences can be used for important applications in livestock, including breed assignment of individual animals, authentication of mono-breed products and parentage verification among several other applications. To identify the most discriminating SNPs among thousands of markers in the available commercial SNP chip tools, several methods have been used. Random forest (RF) is a machine learning technique that has been proposed for this purpose. In this study, we used RF to analyse PorcineSNP60 BeadChip array genotyping data obtained from a total of 2737 pigs of 7 Italian pig breeds (3 cosmopolitan-derived breeds: Italian Large White, Italian Duroc and Italian Landrace, and 4 autochthonous breeds: Apulo-Calabrese, Casertana, Cinta Senese and Nero Siciliano) to identify breed informative and reduced SNP panels using the mean decrease in the Gini Index and the Mean Decrease in Accuracy parameters with stability evaluation. Other reduced informative SNP panels were obtained using Delta, Fixation index and principal component analysis statistics, and their performances were compared with those obtained using the RF-defined panels using the RF classification method and its derived Out Of Bag rates and correct prediction proportions. Therefore, the performances of a total of six reduced panels were evaluated. The correct assignment of the animals to its breed was close to 100% for all tested approaches. Porcine chromosome 8 harboured the largest number of selected SNPs across all panels. Many SNPs were included in genomic regions in which previous studies identified signatures of selection or genes (e.g. ESR1, KITL and LCORL) that could contribute to explain, at least in part, phenotypically or economically relevant traits that might differentiate cosmopolitan and autochthonous pig breeds. Random forest used as preselection statistics highlighted informative SNPs that were not the same as those identified by other methods. This might be due to specific features of this machine learning methodology. It will be interesting to explore if the adaptation of RF methods for the identification of selection signature regions could be able to describe population-specific features that are not captured by other approaches.