Comparison between supervised and unsupervised classifications of neuronal cell types: a case study

Astrocyte ensheathment of calyx-forming axons of the auditory brainstem precedes accelerated expression of myelin genes and myelination by oligodendrocytes

Early postnatal brain development involves complex interactions among maturing neurons and glial cells that drive tissue organization. We previously analyzed gene expression in tissue from the mouse medial nucleus of the trapezoid body (MNTB) during the first postnatal week to study changes that surround rapid growth of the large calyx of Held (CH) nerve terminal. Here, we present genes that show significant changes in gene expression level during the second postnatal week, a developmental timeframe that brackets the onset of airborne sound stimulation and the early stages of myelination. Gene Ontology analysis revealed that many of these genes are related to the myelination process. Further investigation of these genes using a previously published cell type-specific bulk RNA-Seq data set in cortex and our own single-cell RNA-Seq data set in the MNTB revealed enrichment of these genes in the oligodendrocyte lineage (OL) cells. Combining the postnatal day (P)6-P14 microarray gene expression data with the previously published P0-P6 data provided fine temporal resolution to investigate the initiation and subsequent waves of gene expression related to OL cell maturation and the process of myelination. Many genes showed increasing expression levels between P2 and P6 in patterns that reflect OL cell maturation. Correspondingly, the first myelin proteins were detected by P4. Using a complementary, developmental series of electron microscopy 3D image volumes, we analyzed the temporal progression of axon wrapping and myelination in the MNTB. By employing a combination of established ultrastructural criteria to classify reconstructed early postnatal glial cells in the 3D volumes, we demonstrated for the first time that astrocytes within the mouse MNTB extensively wrap the axons of the growing CH terminal prior to OL cell wrapping and compaction of myelin. Our data revealed significant expression of several myelin genes and enrichment of multiple genes associated with lipid metabolism in astrocytes, which may subserve axon wrapping in addition to myelin formation. The transition from axon wrapping by astrocytes to OL cells occurs rapidly between P4 and P9 and identifies a potential new role of astrocytes in priming calyceal axons for subsequent myelination.

DSM: Deep sequential model for complete neuronal morphology representation and feature extraction

The full morphology of single neurons is indispensable for understanding cell types, the basic building blocks in brains. Projecting trajectories are critical to extracting biologically relevant information from neuron morphologies, as they provide valuable information for both connectivity and cell identity. We developed an artificial intelligence method, deep sequential model (DSM), to extract concise, cell-type-defining features from projections across brain regions. DSM achieves more than 90% accuracy in classifying 12 major neuron projection types without compromising performance when spatial noise is present. Such remarkable robustness enabled us to efficiently manage and analyze several major full-morphology data sources, showcasing how characteristic long projections can define cell identities. We also succeeded in applying our model to both discovering previously unknown neuron subtypes and analyzing exceptional co-expressed genes involved in neuron projection circuits.

Topological Sholl descriptors for neuronal clustering and classification

Neuronal morphology is a fundamental factor influencing information processing within neurons and networks. Dendritic morphology in particular can widely vary among cell classes, brain regions, and animal species. Thus, accurate quantitative descriptions allowing classification of large sets of neurons is essential for their structural and functional characterization. Current robust and unbiased computational methods that characterize groups of neurons are scarce. In this work, we introduce a novel technique to study dendritic morphology, complementing and advancing many of the existing techniques. Our approach is to conceptualize the notion of a Sholl descriptor and associate, for each morphological feature, and to each neuron, a function of the radial distance from the soma, taking values in a metric space. Functional distances give rise to pseudo-metrics on sets of neurons which are then used to perform the two distinct tasks of clustering and classification. To illustrate the use of Sholl descriptors, four datasets were retrieved from the large public repository https://neuromorpho.org/ comprising neuronal reconstructions from different species and brain regions. Sholl descriptors were subsequently computed, and standard clustering methods enhanced with detection and metric learning algorithms were then used to objectively cluster and classify each dataset. Importantly, our descriptors outperformed conventional morphometric techniques (L-Measure metrics) in several of the tested datasets. Therefore, we offer a novel and effective approach to the analysis of diverse neuronal cell types, and provide a toolkit for researchers to cluster and classify neurons.

Capturing the songs of mice with an improved detection and classification method for ultrasonic vocalizations (BootSnap)

House mice communicate through ultrasonic vocalizations (USVs), which are above the range of human hearing (>20 kHz), and several automated methods have been developed for USV detection and classification. Here we evaluate their advantages and disadvantages in a full, systematic comparison, while also presenting a new approach. This study aims to 1) determine the most efficient USV detection tool among the existing methods, and 2) develop a classification model that is more generalizable than existing methods. In both cases, we aim to minimize the user intervention required for processing new data. We compared the performance of four detection methods in an out-of-the-box approach, pretrained DeepSqueak detector, MUPET, USVSEG, and the Automatic Mouse Ultrasound Detector (A-MUD). We also compared these methods to human visual or ‘manual’ classification (ground truth) after assessing its reliability. A-MUD and USVSEG outperformed the other methods in terms of true positive rates using default and adjusted settings, respectively, and A-MUD outperformed USVSEG when false detection rates were also considered. For automating the classification of USVs, we developed BootSnap for supervised classification, which combines bootstrapping on Gammatone Spectrograms and Convolutional Neural Networks algorithms with Snapshot ensemble learning. It successfully classified calls into 12 types, including a new class of false positives that is useful for detection refinement. BootSnap outperformed the pretrained and retrained state-of-the-art tool, and thus it is more generalizable. BootSnap is freely available for scientific use.

House mice and many other species use ultrasonic vocalizations to communicate in various contexts including social and sexual interactions. These vocalizations are increasingly investigated in research on animal communication and as a phenotype for studying the genetic basis of autism and speech disorders. Because manual methods for analyzing vocalizations are extremely time consuming, automatic tools for detection and classification are needed. We evaluated the performance of the available tools for analyzing ultrasonic vocalizations, and we compared detection tools for the first time to manual methods (“ground truth”) using recordings from wild-derived and laboratory mice. For the first time, class-wise inter-observer reliability of manual labels used for ground truth are analyzed and reported. Moreover, we developed a new classification method based on ensemble deep learning that provides more generalizability than the current state-of-the-art tool (both pretrained and retrained). Our new classification method is free for scientific use.

Comparison of Different Machine Learning Techniques to Predict Diabetic Kidney Disease

Background

Diabetic kidney disease (DKD), one of the complications of diabetes in patients, leads to progressive loss of kidney function. Timely intervention is known to improve outcomes. Therefore, screening patients to identify high-risk populations is important. Machine learning classification techniques can be applied to patient datasets to identify high-risk patients by building a predictive model.

Objective

This study aims to identify a suitable classification technique for predicting DKD by applying different classification techniques to a DKD dataset and comparing their performance using WEKA machine learning software.

Methods

The performance of nine different classification techniques was analyzed on a DKD dataset with 410 instances and 18 attributes. Data preprocessing was carried out using the PartitionMembershipFilter. A 10-fold cross validation was performed on the dataset. The performance was assessed on the basis of the execution time, accuracy, correctly and incorrectly classified instances, kappa statistics (K), mean absolute error, root mean squared error, and true values of the confusion matrix.

Results

With an accuracy of 93.6585% and a higher K value (0.8731), IBK and random tree classification techniques were found to be the best performing techniques. Moreover, they also exhibited the lowest root mean squared error rate (0.2496). There were 15 false-positive instances and 11 false-negative instances with these prediction models.

Conclusions

This study identified IBK and random tree classification techniques as the best performing classifiers and accurate prediction methods for DKD.

Volatile fingerprint of food products with untargeted SIFT-MS data coupled with mixOmics methods for profile discrimination: Application case on cheese

Volatile organic compounds (VOCs) emitted by food products are decisive for the perception of aroma and taste. The analysis of gaseous matrices is traditionally done by detection and quantification of few dozens of characteristic markers. Emerging direct injection mass spectrometry technologies offer rapid analysis based on a soft ionisation of VOCs without previous separation. The recent increase of selectivity offered by the use of several precursor ions coupled with untargeted analysis increases the potential power of these instruments. However, the analysis of complex gaseous matrix results in a large number of ion conflicts, making the quantification of markers difficult, and in a large volume of data. In this work, we present the exploitation of untargeted SIFT-MS volatile fingerprints of ewe PDO cheeses in a real farm model, using mixOmics methods allowing us to illustrate the typicality, the manufacturing processes reproducibility and the impact of the animals' diet on the final product.

Morpho-physiological properties and connectivity of vasoactive intestinal polypeptide-expressing interneurons in the mouse hippocampal dentate gyrus

The hippocampus is a key brain structure for cognitive and emotional functions. Among the hippocampal subregions, the dentate gyrus (DG) is the first station that receives multimodal sensory information from the cortex. Local-circuit inhibitory GABAergic interneurons (INs) regulate the excitation-inhibition balance in the DG principal neurons (PNs) and therefore are critical for information processing. Similar to PNs, GABAergic INs also receive distinct inhibitory inputs. Among various classes of INs, vasoactive intestinal polypeptide-expressing (VIP⁺ ) INs preferentially target other INs in several brain regions and thereby directly modulate the GABAergic system. However, the morpho-physiological characteristics and postsynaptic targets of VIP⁺ INs in the DG are poorly understood. Here, we report that VIP⁺ INs in the mouse DG are highly heterogeneous based on their morpho-physiological characteristics. In approximately two-thirds of morphologically reconstructed cells, their axons ramify in the hilus. The remaining cells project their axons exclusively to the molecular layer (15%), to both the molecular layer and hilus (10%), or throughout the entire DG layers (8%). Generally, VIP⁺ INs display variable intrinsic properties and discharge patterns without clear correlation with their morphologies. Finally, VIP⁺ INs are recruited with a long latency in response to theta-band cortical inputs and preferentially innervate GABAergic INs over glutamatergic PNs. In summary, VIP⁺ INs in the DG are composed of highly diverse subpopulations and control the DG output via disinhibition.

A community-based transcriptomics classification and nomenclature of neocortical cell types

To understand the function of cortical circuits, it is necessary to catalog their cellular diversity. Past attempts to do so using anatomical, physiological or molecular features of cortical cells have not resulted in a unified taxonomy of neuronal or glial cell types, partly due to limited data. Single-cell transcriptomics is enabling, for the first time, systematic high-throughput measurements of cortical cells and generation of datasets that hold the promise of being complete, accurate and permanent. Statistical analyses of these data reveal clusters that often correspond to cell types previously defined by morphological or physiological criteria and that appear conserved across cortical areas and species. To capitalize on these new methods, we propose the adoption of a transcriptome-based taxonomy of cell types for mammalian neocortex. This classification should be hierarchical and use a standardized nomenclature. It should be based on a probabilistic definition of a cell type and incorporate data from different approaches, developmental stages and species. A community-based classification and data aggregation model, such as a knowledge graph, could provide a common foundation for the study of cortical circuits. This community-based classification, nomenclature and data aggregation could serve as an example for cell type atlases in other parts of the body.

Classification, Categorization, and Algorithms for Articular Cartilage Defects

There is a critical unmet need in the clinical implementation of valid preventative and therapeutic strategies for patients with articular cartilage pathology based on the significant gap in understanding of the relationships between diagnostic data, disease progression, patient-related variables, and symptoms. In this article, the current state of classification and categorization for articular cartilage pathology is discussed with particular focus on machine learning methods and the authors propose a bedside-bench-bedside approach with highly quantitative techniques as a solution to these hurdles. Leveraging computational learning with available data toward articular cartilage pathology patient phenotyping holds promise for clinical research and will likely be an important tool to identify translational solutions into evidence-based clinical applications to benefit patients. Recommendations for successful implementation of these approaches include using standardized definitions of articular cartilage, to include characterization of depth, size, location, and number; using measurements that minimize subjectivity or validated patient-reported outcome measures; considering not just the articular cartilage pathology but the whole joint, and the patient perception and perspective. Application of this approach through a multistep process by a multidisciplinary team of clinicians and scientists holds promise for validating disease mechanism-based phenotypes toward clinically relevant understanding of articular cartilage pathology for evidence-based application to orthopaedic practice.

New morphological features based on the Sholl analysis for automatic classification of traced neurons

BACKGROUND

This article addresses the automatic classification of reconstructed neurons through their morphological features. The purpose was to extend the capabilities of the L-Measure software.

METHODS

New morphological features were developed, based on modifications of the conventional Sholl analysis. The lengths of the compartments, as well as their volumes, were added to the features used in the classical analysis in order to improve the results during automatic neuron classification. FSM were used to obtain subsets of lower cardinality from the full feature sets and the usefulness of these subsets was tested through their use in supervised classification tasks. The study was based on two types of neurons belonging to mice: pyramidal and GABAergic interneurons. Furthermore, a set of pyramidal neurons belonging to Later 4 and Layer 5 was analyzed.

RESULTS

RF classifier shown the best performance combined with a Wrapper method.U-WNAD set allowed to obtain higher values than WN, A and D in all cases and better results than LM for the filters and wrappers FSM. U-LM-WNAD set, led to the highest AUC values for all the FSM studied. Similar results for different regions of cortex were obtained. Comparison with Existing Methods The new features exhibited high discriminatory power with which the values of AUC and Acc obtained in the experiments exceeded those obtained using only the features provided by L-Measure.

CONCLUSIONS

The highest values of AUC and Acc were obtained from the sets U-WNAD and U-LM-WNAD, evidencing the discriminatory power of the new proposed features.

Measurements of neuronal morphological variation across the rat neocortex

Neuron morphology is highly variable across the mammalian brain. It is thought that these attributes of neuronal cell shape, such as soma surface area and branching frequency, are determined by biological function and information processing. In this study, a large data set of neurons across the rat neocortex were clustered by their anatomical characters for evidence of distinctiveness among neocortical regions and the somatosensory layers. This data set of neuronal morphologies was compiled from 31 different lab sources with a validation procedure so that data records are potentially comparable across research studies. With this large set of heterogeneous data and by clustering analysis, this study shows that neuronal morphological traits overlap among neocortical and somatosensory regions. In the context of past neuroanatomical studies, this result is not congruent with tissue level analysis and strongly suggests further sampling of neuronal data to lessen the effect of confounding factors, such as the influence of different methodologies from use of heterogeneous samples of neuronal data.

Molecular profiling of single neurons of known identity in two ganglia from the crab Cancer borealis

Single-cell transcriptional profiling has become a widespread tool in cell identification, particularly in the nervous system, based on the notion that genomic information determines cell identity. However, many cell-type classification studies are unconstrained by other cellular attributes (e.g., morphology, physiology). Here, we systematically test how accurately transcriptional profiling can assign cell identity to well-studied anatomically and functionally identified neurons in 2 small neuronal networks. While these neurons clearly possess distinct patterns of gene expression across cell types, their expression profiles are not sufficient to unambiguously confirm their identity. We suggest that true cell identity can only be determined by combining gene expression data with other cellular attributes such as innervation pattern, morphology, or physiology.

Understanding circuit organization depends on identification of cell types. Recent advances in transcriptional profiling methods have enabled classification of cell types by their gene expression. While exceptionally powerful and high throughput, the ground-truth validation of these methods is difficult: If cell type is unknown, how does one assess whether a given analysis accurately captures neuronal identity? To shed light on the capabilities and limitations of solely using transcriptional profiling for cell-type classification, we performed 2 forms of transcriptional profiling—RNA-seq and quantitative RT-PCR, in single, unambiguously identified neurons from 2 small crustacean neuronal networks: The stomatogastric and cardiac ganglia. We then combined our knowledge of cell type with unbiased clustering analyses and supervised machine learning to determine how accurately functionally defined neuron types can be classified by expression profile alone. The results demonstrate that expression profile is able to capture neuronal identity most accurately when combined with multimodal information that allows for post hoc grouping, so analysis can proceed from a supervised perspective. Solely unsupervised clustering can lead to misidentification and an inability to distinguish between 2 or more cell types. Therefore, this study supports the general utility of cell identification by transcriptional profiling, but adds a caution: It is difficult or impossible to know under what conditions transcriptional profiling alone is capable of assigning cell identity. Only by combining multiple modalities of information such as physiology, morphology, or innervation target can neuronal identity be unambiguously determined.

Neuronal Morphology and Synapse Count in the Nematode Worm

The somatic nervous system of the nematode worm Caenorhabditis elegans is a model for understanding the physical characteristics of the neurons and their interconnections. Its neurons show high variation in morphological attributes. This study investigates the relationship of neuronal morphology to the number of synapses per neuron. Morphology is also examined for any detectable association with neuron cell type or ganglion membership.

Alternative classifications of neurons based on physiological properties and synaptic responses, a computational study

One of the central goals of today's neuroscience is to achieve the conceivably most accurate classification of neuron types in the mammalian brain. As part of this research effort, electrophysiologists commonly utilize current clamp techniques to gain a detailed characterization of the neurons' physiological properties. While this approach has been useful, it is not well understood whether neurons that share physiological properties of a particular phenotype would also operate consistently under the action of natural synaptic inputs. We approached this problem by simulating a biophysically diverse population of model neurons based on 3 generic phenotypes. We exposed the model neurons to two types of stimulation to investigate their voltage responses under conventional current step protocols and under simulated synaptic bombardment. We extracted standard physiological parameters from the voltage responses elicited by current step stimulation and spike arrival times descriptive of the model's firing behavior under synaptic inputs. The biophysical phenotypes could be reliably identified using classification based on the 'static' physiological properties, but not the interspike interval-based parameters. However, the model neurons associated with the biophysically different phenotypes retained cell type specific features in the fine structure of their spike responses that allowed their accurate classification.

Evaluation of Machine Learning Algorithms for Surface Water Extraction in a Landsat 8 Scene of Nepal

With over 6000 rivers and 5358 lakes, surface water is one of the most important resources in Nepal. However, the quantity and quality of Nepal's rivers and lakes are decreasing due to human activities and climate change. Despite the advancement of remote sensing technology and the availability of open access data and tools, the monitoring and surface water extraction works has not been carried out in Nepal. Single or multiple water index methods have been applied in the extraction of surface water with satisfactory results. Extending our previous study, the authors evaluated six different machine learning algorithms: Naive Bayes (NB), recursive partitioning and regression trees (RPART), neural networks (NNET), support vector machines (SVM), random forest (RF), and gradient boosted machines (GBM) to extract surface water in Nepal. With three secondary bands, slope, NDVI and NDWI, the algorithms were evaluated for performance with the addition of extra information. As a result, all the applied machine learning algorithms, except NB and RPART, showed good performance. RF showed overall accuracy (OA) and kappa coefficient (Kappa) of 1 for the all the multiband data with the reference dataset, followed by GBM, NNET, and SVM in metrics. The performances were better in the hilly regions and flat lands, but not well in the Himalayas with ice, snow and shadows, and the addition of slope and NDWI showed improvement in the results. Adding single secondary bands is better than adding multiple in most algorithms except NNET. From current and previous studies, it is recommended to separate any study area with and without snow or low and high elevation, then apply machine learning algorithms in original Landsat data or with the addition of slopes or NDWI for better performance.

Towards a supervised classification of neocortical interneuron morphologies

BACKGROUND

The challenge of classifying cortical interneurons is yet to be solved. Data-driven classification into established morphological types may provide insight and practical value.

RESULTS

We trained models using 217 high-quality morphologies of rat somatosensory neocortex interneurons reconstructed by a single laboratory and pre-classified into eight types. We quantified 103 axonal and dendritic morphometrics, including novel ones that capture features such as arbor orientation, extent in layer one, and dendritic polarity. We trained a one-versus-rest classifier for each type, combining well-known supervised classification algorithms with feature selection and over- and under-sampling. We accurately classified the nest basket, Martinotti, and basket cell types with the Martinotti model outperforming 39 out of 42 leading neuroscientists. We had moderate accuracy for the double bouquet, small and large basket types, and limited accuracy for the chandelier and bitufted types. We characterized the types with interpretable models or with up to ten morphometrics.

CONCLUSION

Except for large basket, 50 high-quality reconstructions sufficed to learn an accurate model of a type. Improving these models may require quantifying complex arborization patterns and finding correlates of bouton-related features. Our study brings attention to practical aspects important for neuron classification and is readily reproducible, with all code and data available online.

Electrophysiological Profiling of Neocortical Neural Subtypes: A Semi-Supervised Method Applied to in vivo Whole-Cell Patch-Clamp Data

A lot of efforts have been made to understand the structure and function of neocortical circuits. In fact, a promising way to understand the functions of cortical circuits is the classification of the neural types, based on their different properties. Recent studies focused on applying modern computational methods to classify neurons based on molecular, morphological, physiological, or mixed of these criteria. Although there are studies in the literature on in vitro/vivo extracellular or in vitro intracellular recordings, a study on the classification of neuronal types using in vivo whole-cell patch-clamp recordings is still lacking. We thus proposed a novel semi-supervised classification method based on waveform shape of neurons' spikes using in vivo whole-cell patch-clamp recordings. We, first, detected spike candidates. Then discriminative features were extracted from the time samples of the spikes using discrete cosine transform. We then extracted the center of clusters using fuzzy c-mean clustering and finally, the neurons were classified using the minimum distance classifier. We distinguished three types of neurons: excitatory pyramidal cells (Pyr) and two types of inhibitory neurons: GABAergic- parvalbumin positive (PV), and somatostatin positive (SST) non-pyramidal cells in layer II/III of the mice primary visual cortex. We used 10-fold cross validation in our study. The classification accuracy for PV, Pyr, and SST was 91.59 ± 1.69, 97.47 ± 0.67, and 89.06 ± 1.99, respectively. Overall, the algorithm correctly classified 92.67 ± 0.54% of the cells, confirming the relative robustness of the discriminant functions. The performance of the method was further assessed on in vitro recordings by using a pool of 50 neurons from Allen institute Cell Types Database (5 major subtypes of neurons: Pyr, PV, SST, 5HT3a, and vasoactive intestinal peptide (VIP) cells). Its overall accuracy was 84.13 ± 0.81% on this data set using cross validation framework. The proposed algorithm is thus a promising new tool in recognizing cell's type with high accuracy in laboratories using in vivo/vitro whole-cell patch-clamp recording technique. The developed programs and the entire dataset are available online to interested readers.

Detection and Classification of Overlapping Cell Nuclei in Cytology Effusion Images Using a Double-Strategy Random Forest

Due to the close resemblance between overlapping and cancerous nuclei, the misinterpretation of overlapping nuclei can affect the final decision of cancer cell detection. Thus, it is essential to detect overlapping nuclei and distinguish them from single ones for subsequent quantitative analyses. This paper presents a method for the automated detection and classification of overlapping nuclei from single nuclei appearing in cytology pleural effusion (CPE) images. The proposed system is comprised of three steps: nuclei candidate extraction, dominant feature extraction, and classification of single and overlapping nuclei. A maximum entropy thresholding method complemented by image enhancement and post-processing was employed for nuclei candidate extraction. For feature extraction, a new combination of 16 geometrical and 10 textural features was extracted from each nucleus region. A double-strategy random forest was performed as an ensemble feature selector to select the most relevant features, and an ensemble classifier to differentiate between overlapping nuclei and single ones using selected features. The proposed method was evaluated on 4000 nuclei from CPE images using various performance metrics. The results were 96.6% sensitivity, 98.7% specificity, 92.7% precision, 94.6% F1 score, 98.4% accuracy, 97.6% G-mean, and 99% area under curve. The computation time required to run the entire algorithm was just 5.17 s. The experiment results demonstrate that the proposed algorithm yields a superior performance to previous studies and other classifiers. The proposed algorithm can serve as a new supportive tool in the automated diagnosis of cancer cells from cytology images.

Mining Big Neuron Morphological Data

The advent of automatic tracing and reconstruction technology has led to a surge in the number of neurons 3D reconstruction data and consequently the neuromorphology research. However, the lack of machine-driven annotation schema to automatically detect the types of the neurons based on their morphology still hinders the development of this branch of science. Neuromorphology is important because of the interplay between the shape and functionality of neurons and the far-reaching impact on the diagnostics and therapeutics in neurological disorders. This survey paper provides a comprehensive research in the field of automatic neurons classification and presents the existing challenges, methods, tools, and future directions for automatic neuromorphology analytics. We summarize the major automatic techniques applicable in the field and propose a systematic data processing pipeline for automatic neuron classification, covering data capturing, preprocessing, analyzing, classification, and retrieval. Various techniques and algorithms in machine learning are illustrated and compared to the same dataset to facilitate ongoing research in the field.

New Features for Neuron Classification

This paper addresses the problem of obtaining new neuron features capable of improving results of neuron classification. Most studies on neuron classification using morphological features have been based on Euclidean geometry. Here three one-dimensional (1D) time series are derived from the three-dimensional (3D) structure of neuron instead, and afterwards a spatial time series is finally constructed from which the features are calculated. Digitally reconstructed neurons were separated into control and pathological sets, which are related to three categories of alterations caused by epilepsy, Alzheimer's disease (long and local projections), and ischemia. These neuron sets were then subjected to supervised classification and the results were compared considering three sets of features: morphological, features obtained from the time series and a combination of both. The best results were obtained using features from the time series, which outperformed the classification using only morphological features, showing higher correct classification rates with differences of 5.15, 3.75, 5.33% for epilepsy and Alzheimer's disease (long and local projections) respectively. The morphological features were better for the ischemia set with a difference of 3.05%. Features like variance, Spearman auto-correlation, partial auto-correlation, mutual information, local minima and maxima, all related to the time series, exhibited the best performance. Also we compared different evaluators, among which ReliefF was the best ranked.

Distinct Functional Groups Emerge from the Intrinsic Properties of Molecularly Identified Entorhinal Interneurons and Principal Cells

Inhibitory interneurons are an important source of synaptic inputs that may contribute to network mechanisms for coding of spatial location by entorhinal cortex (EC). The intrinsic properties of inhibitory interneurons in the EC of the mouse are mostly undescribed. Intrinsic properties were recorded from known cell types, such as, stellate and pyramidal cells and 6 classes of molecularly identified interneurons (regulator of calcineurin 2, somatostatin, serotonin receptor 3a, neuropeptide Y neurogliaform (NGF), neuropeptide Y non-NGF, and vasoactive intestinal protein) in acute brain slices. We report a broad physiological diversity between and within cell classes. We also found differences in the ability to produce postinhibitory rebound spikes and in the frequency and amplitude of incoming EPSPs. To understand the source of this intrinsic variability we applied hierarchical cluster analysis to functionally classify neurons. These analyses revealed physiologically derived cell types in EC that mostly corresponded to the lines identified by biomarkers with a few unexpected and important differences. Finally, we reduced the complex multidimensional space of intrinsic properties to the most salient five that predicted the cellular biomolecular identity with 81.4% accuracy. These results provide a framework for the classification of functional subtypes of cortical neurons by their intrinsic membrane properties.

Metrics for comparing neuronal tree shapes based on persistent homology

As more and more neuroanatomical data are made available through efforts such as NeuroMorpho.Org and FlyCircuit.org, the need to develop computational tools to facilitate automatic knowledge discovery from such large datasets becomes more urgent. One fundamental question is how best to compare neuron structures, for instance to organize and classify large collection of neurons. We aim to develop a flexible yet powerful framework to support comparison and classification of large collection of neuron structures efficiently. Specifically we propose to use a topological persistence-based feature vectorization framework. Existing methods to vectorize a neuron (i.e, convert a neuron to a feature vector so as to support efficient comparison and/or searching) typically rely on statistics or summaries of morphometric information, such as the average or maximum local torque angle or partition asymmetry. These simple summaries have limited power in encoding global tree structures. Based on the concept of topological persistence recently developed in the field of computational topology, we vectorize each neuron structure into a simple yet informative summary. In particular, each type of information of interest can be represented as a descriptor function defined on the neuron tree, which is then mapped to a simple persistence-signature. Our framework can encode both local and global tree structure, as well as other information of interest (electrophysiological or dynamical measures), by considering multiple descriptor functions on the neuron. The resulting persistence-based signature is potentially more informative than simple statistical summaries (such as average/mean/max) of morphometric quantities-Indeed, we show that using a certain descriptor function will give a persistence-based signature containing strictly more information than the classical Sholl analysis. At the same time, our framework retains the efficiency associated with treating neurons as points in a simple Euclidean feature space, which would be important for constructing efficient searching or indexing structures over them. We present preliminary experimental results to demonstrate the effectiveness of our persistence-based neuronal feature vectorization framework.

Morphological Neuron Classification Using Machine Learning

Classification and quantitative characterization of neuronal morphologies from histological neuronal reconstruction is challenging since it is still unclear how to delineate a neuronal cell class and which are the best features to define them by. The morphological neuron characterization represents a primary source to address anatomical comparisons, morphometric analysis of cells, or brain modeling. The objectives of this paper are (i) to develop and integrate a pipeline that goes from morphological feature extraction to classification and (ii) to assess and compare the accuracy of machine learning algorithms to classify neuron morphologies. The algorithms were trained on 430 digitally reconstructed neurons subjectively classified into layers and/or m-types using young and/or adult development state population of the somatosensory cortex in rats. For supervised algorithms, linear discriminant analysis provided better classification results in comparison with others. For unsupervised algorithms, the affinity propagation and the Ward algorithms provided slightly better results.

Unbiased Mitoproteome Analyses Confirm Non-canonical RNA, Expanded Codon Translations

Proteomic MS/MS mass spectrometry detections are usually biased towards peptides cleaved by experimentally added digestion enzyme(s). Hence peptides resulting from spontaneous degradation and natural proteolysis usually remain undetected. Previous analyses of tryptic human proteome data (cleavage after K, R) detected non-canonical tryptic peptides translated according to tetra- and pentacodons (codons expanded by silent mono- and dinucleotides), and from transcripts systematically (a) deleting mono-, dinucleotides after trinucleotides (delRNAs), (b) exchanging nucleotides according to 23 bijective transformations. Nine symmetric and fourteen asymmetric nucleotide exchanges (X ↔ Y, e.g. A ↔ C; and X → Y → Z → X, e.g. A → C → G → A) produce swinger RNAs. Here unbiased reanalyses of these proteomic data detect preferentially non-canonical tryptic peptides despite assuming random cleavage. Unbiased analyses couldn't reconstruct experimental tryptic digestion if most detected non-canonical peptides were false positives. Detected non-tryptic non-canonical peptides map preferentially on corresponding, previously described non-canonical transcripts, as for tryptic non-canonical peptides. Hence unbiased analyses independently confirm previous trypsin-biased analyses that showed translations of del- and swinger RNA and expanded codons. Accounting for natural proteolysis completes trypsin-biased mitopeptidome analyses, independently confirms non-canonical transcriptions and translations.

Automatic discovery of cell types and microcircuitry from neural connectomics

Neural connectomics has begun producing massive amounts of data, necessitating new analysis methods to discover the biological and computational structure. It has long been assumed that discovering neuron types and their relation to microcircuitry is crucial to understanding neural function. Here we developed a non-parametric Bayesian technique that identifies neuron types and microcircuitry patterns in connectomics data. It combines the information traditionally used by biologists in a principled and probabilistically coherent manner, including connectivity, cell body location, and the spatial distribution of synapses. We show that the approach recovers known neuron types in the retina and enables predictions of connectivity, better than simpler algorithms. It also can reveal interesting structure in the nervous system of Caenorhabditis elegans and an old man-made microprocessor. Our approach extracts structural meaning from connectomics, enabling new approaches of automatically deriving anatomical insights from these emerging datasets.

Towards the automatic classification of neurons

The classification of neurons into types has been much debated since the inception of modern neuroscience. Recent experimental advances are accelerating the pace of data collection. The resulting growth of information about morphological, physiological, and molecular properties encourages efforts to automate neuronal classification by powerful machine learning techniques. We review state-of-the-art analysis approaches and the availability of suitable data and resources, highlighting prominent challenges and opportunities. The effective solution of the neuronal classification problem will require continuous development of computational methods, high-throughput data production, and systematic metadata organization to enable cross-laboratory integration.

Classifying GABAergic interneurons with semi-supervised projected model-based clustering

OBJECTIVES

A recently introduced pragmatic scheme promises to be a useful catalog of interneuron names. We sought to automatically classify digitally reconstructed interneuronal morphologies according to this scheme. Simultaneously, we sought to discover possible subtypes of these types that might emerge during automatic classification (clustering). We also investigated which morphometric properties were most relevant for this classification.

MATERIALS AND METHODS

A set of 118 digitally reconstructed interneuronal morphologies classified into the common basket (CB), horse-tail (HT), large basket (LB), and Martinotti (MA) interneuron types by 42 of the world's leading neuroscientists, quantified by five simple morphometric properties of the axon and four of the dendrites. We labeled each neuron with the type most commonly assigned to it by the experts. We then removed this class information for each type separately, and applied semi-supervised clustering to those cells (keeping the others' cluster membership fixed), to assess separation from other types and look for the formation of new groups (subtypes). We performed this same experiment unlabeling the cells of two types at a time, and of half the cells of a single type at a time. The clustering model is a finite mixture of Gaussians which we adapted for the estimation of local (per-cluster) feature relevance. We performed the described experiments on three different subsets of the data, formed according to how many experts agreed on type membership: at least 18 experts (the full data set), at least 21 (73 neurons), and at least 26 (47 neurons).

RESULTS

Interneurons with more reliable type labels were classified more accurately. We classified HT cells with 100% accuracy, MA cells with 73% accuracy, and CB and LB cells with 56% and 58% accuracy, respectively. We identified three subtypes of the MA type, one subtype of CB and LB types each, and no subtypes of HT (it was a single, homogeneous type). We got maximum (adapted) Silhouette width and ARI values of 1, 0.83, 0.79, and 0.42, when unlabeling the HT, CB, LB, and MA types, respectively, confirming the quality of the formed cluster solutions. The subtypes identified when unlabeling a single type also emerged when unlabeling two types at a time, confirming their validity. Axonal morphometric properties were more relevant that dendritic ones, with the axonal polar histogram length in the [π, 2π) angle interval being particularly useful.

CONCLUSIONS

The applied semi-supervised clustering method can accurately discriminate among CB, HT, LB, and MA interneuron types while discovering potential subtypes, and is therefore useful for neuronal classification. The discovery of potential subtypes suggests that some of these types are more heterogeneous that previously thought. Finally, axonal variables seem to be more relevant than dendritic ones for distinguishing among the CB, HT, LB, and MA interneuron types.

Multi-dimensional classification of GABAergic interneurons with Bayesian network-modeled label uncertainty

Interneuron classification is an important and long-debated topic in neuroscience. A recent study provided a data set of digitally reconstructed interneurons classified by 42 leading neuroscientists according to a pragmatic classification scheme composed of five categorical variables, namely, of the interneuron type and four features of axonal morphology. From this data set we now learned a model which can classify interneurons, on the basis of their axonal morphometric parameters, into these five descriptive variables simultaneously. Because of differences in opinion among the neuroscientists, especially regarding neuronal type, for many interneurons we lacked a unique, agreed-upon classification, which we could use to guide model learning. Instead, we guided model learning with a probability distribution over the neuronal type and the axonal features, obtained, for each interneuron, from the neuroscientists' classification choices. We conveniently encoded such probability distributions with Bayesian networks, calling them label Bayesian networks (LBNs), and developed a method to predict them. This method predicts an LBN by forming a probabilistic consensus among the LBNs of the interneurons most similar to the one being classified. We used 18 axonal morphometric parameters as predictor variables, 13 of which we introduce in this paper as quantitative counterparts to the categorical axonal features. We were able to accurately predict interneuronal LBNs. Furthermore, when extracting crisp (i.e., non-probabilistic) predictions from the predicted LBNs, our method outperformed related work on interneuron classification. Our results indicate that our method is adequate for multi-dimensional classification of interneurons with probabilistic labels. Moreover, the introduced morphometric parameters are good predictors of interneuron type and the four features of axonal morphology and thus may serve as objective counterparts to the subjective, categorical axonal features.

Bayesian networks in neuroscience: a survey

Bayesian networks are a type of probabilistic graphical models lie at the intersection between statistics and machine learning. They have been shown to be powerful tools to encode dependence relationships among the variables of a domain under uncertainty. Thanks to their generality, Bayesian networks can accommodate continuous and discrete variables, as well as temporal processes. In this paper we review Bayesian networks and how they can be learned automatically from data by means of structure learning algorithms. Also, we examine how a user can take advantage of these networks for reasoning by exact or approximate inference algorithms that propagate the given evidence through the graphical structure. Despite their applicability in many fields, they have been little used in neuroscience, where they have focused on specific problems, like functional connectivity analysis from neuroimaging data. Here we survey key research in neuroscience where Bayesian networks have been used with different aims: discover associations between variables, perform probabilistic reasoning over the model, and classify new observations with and without supervision. The networks are learned from data of any kind-morphological, electrophysiological, -omics and neuroimaging-, thereby broadening the scope-molecular, cellular, structural, functional, cognitive and medical- of the brain aspects to be studied.

Synaptic and cellular organization of layer 1 of the developing rat somatosensory cortex

Layer 1 of the neocortex is sparsely populated with neurons and heavily innervated by fibers from lower layers and proximal and distal brain regions. Understanding the potential functions of this layer requires a comprehensive understanding of its cellular and synaptic organization. We therefore performed a quantitative study of the microcircuitry of neocortical layer 1 (L1) in the somatosensory cortex in juvenile rats (P13–P16) using multi-neuron patch-clamp and 3D morphology reconstructions. Expert-based subjective classification of the morphologies of the recorded L1 neurons suggest 6 morphological classes: (1) the Neurogliaform cells with dense axonal arborizations (NGC-DA) and with sparse arborizations (NGC-SA), (2) the Horizontal Axon Cell (HAC), (3) those with descending axonal collaterals (DAC), (4) the large axon cell (LAC), and (5) the small axon cell (SAC). Objective, supervised and unsupervised cluster analyses confirmed DAC, HAC, LAC and NGC as distinct morphological classes. The neurons were also classified into 5 electrophysiological types based on the Petilla convention; classical non-adapting (cNAC), burst non-adapting (bNAC), classical adapting (cAC), classical stuttering (cSTUT), and classical irregular spiking (cIR). The most common electrophysiological type of neuron was the cNAC type (40%) and the most common morpho-electrical type was the NGC-DA—cNAC. Paired patch-clamp recordings revealed that the neurons were connected via GABAergic inhibitory synaptic connections with a 7.9% connection probability and via gap junctions with a 5.2% connection probability. Most synaptic connections were mediated by both GABA_A and GABA_B receptors (62.6%). A smaller fraction of synaptic connections were mediated exclusively by GABA_A (15.4%) or GABA_B (21.8%) receptors. Morphological 3D reconstruction of synaptic connected pairs of L1 neurons revealed multi-synapse connections with an average of 9 putative synapses per connection. These putative synapses were widely distributed with 39% on somata and 61% on dendrites. We also discuss the functional implications of this L1 cellular and synaptic organization in neocortical information processing.

Classification of neocortical interneurons using affinity propagation

In spite of over a century of research on cortical circuits, it is still unknown how many classes of cortical neurons exist. In fact, neuronal classification is a difficult problem because it is unclear how to designate a neuronal cell class and what are the best characteristics to define them. Recently, unsupervised classifications using cluster analysis based on morphological, physiological, or molecular characteristics, have provided quantitative and unbiased identification of distinct neuronal subtypes, when applied to selected datasets. However, better and more robust classification methods are needed for increasingly complex and larger datasets. Here, we explored the use of affinity propagation, a recently developed unsupervised classification algorithm imported from machine learning, which gives a representative example or exemplar for each cluster. As a case study, we applied affinity propagation to a test dataset of 337 interneurons belonging to four subtypes, previously identified based on morphological and physiological characteristics. We found that affinity propagation correctly classified most of the neurons in a blind, non-supervised manner. Affinity propagation outperformed Ward's method, a current standard clustering approach, in classifying the neurons into 4 subtypes. Affinity propagation could therefore be used in future studies to validly classify neurons, as a first step to help reverse engineer neural circuits.

HCS-Neurons: identifying phenotypic changes in multi-neuron images upon drug treatments of high-content screening

Background

High-content screening (HCS) has become a powerful tool for drug discovery. However, the discovery of drugs targeting neurons is still hampered by the inability to accurately identify and quantify the phenotypic changes of multiple neurons in a single image (named multi-neuron image) of a high-content screen. Therefore, it is desirable to develop an automated image analysis method for analyzing multi-neuron images.

Results

We propose an automated analysis method with novel descriptors of neuromorphology features for analyzing HCS-based multi-neuron images, called HCS-neurons. To observe multiple phenotypic changes of neurons, we propose two kinds of descriptors which are neuron feature descriptor (NFD) of 13 neuromorphology features, e.g., neurite length, and generic feature descriptors (GFDs), e.g., Haralick texture. HCS-neurons can 1) automatically extract all quantitative phenotype features in both NFD and GFDs, 2) identify statistically significant phenotypic changes upon drug treatments using ANOVA and regression analysis, and 3) generate an accurate classifier to group neurons treated by different drug concentrations using support vector machine and an intelligent feature selection method. To evaluate HCS-neurons, we treated P19 neurons with nocodazole (a microtubule depolymerizing drug which has been shown to impair neurite development) at six concentrations ranging from 0 to 1000 ng/mL. The experimental results show that all the 13 features of NFD have statistically significant difference with respect to changes in various levels of nocodazole drug concentrations (NDC) and the phenotypic changes of neurites were consistent to the known effect of nocodazole in promoting neurite retraction. Three identified features, total neurite length, average neurite length, and average neurite area were able to achieve an independent test accuracy of 90.28% for the six-dosage classification problem. This NFD module and neuron image datasets are provided as a freely downloadable MatLab project at http://iclab.life.nctu.edu.tw/HCS-Neurons.

Conclusions

Few automatic methods focus on analyzing multi-neuron images collected from HCS used in drug discovery. We provided an automatic HCS-based method for generating accurate classifiers to classify neurons based on their phenotypic changes upon drug treatments. The proposed HCS-neurons method is helpful in identifying and classifying chemical or biological molecules that alter the morphology of a group of neurons in HCS.

Subgroups of parvalbumin-expressing interneurons in layers 2/3 of the visual cortex

The input, processing, and output characteristics of inhibitory interneurons help shape information flow through layers 2/3 of the visual cortex. Parvalbumin (PV)-positive interneurons modulate and synchronize the gain and dynamic responsiveness of pyramidal neurons. To define the diversity of PV interneurons in layers 2/3 of the developing visual cortex, we characterized their passive and active membrane properties. Using Ward's and k-means multidimensional clustering, we identified four PV interneuron subgroups. The most notable difference between the subgroups was their firing patterns in response to moderate stimuli just above rheobase. Two subgroups showed regular and continuous firing at all stimulus intensities above rheobase. The difference between these two continuously firing subgroups was that one fired at much higher frequencies and transitioned into this high-frequency firing rate at or near rheobase. The two other subgroups showed irregular, stuttering firing patterns just above rheobase. Both of these subgroups typically transitioned to regular and continuous firing at intense stimulations, but one of these subgroups, the strongly stuttering subgroup, showed irregular firing across a wider range of stimulus intensities and firing frequencies. The four subgroups also differed in excitatory synaptic input, providing independent support for the classification of subgroups. The subgroups of PV interneurons identified here would respond differently to inputs of varying intensity and frequency, generating diverse patterns of PV inhibition in the developing neural circuit.

A hierarchical structure of cortical interneuron electrical diversity revealed by automated statistical analysis

Although the diversity of cortical interneuron electrical properties is well recognized, the number of distinct electrical types (e-types) is still a matter of debate. Recently, descriptions of interneuron variability were standardized by multiple laboratories on the basis of a subjective classification scheme as set out by the Petilla convention (Petilla Interneuron Nomenclature Group, PING). Here, we present a quantitative, statistical analysis of a database of nearly five hundred neurons manually annotated according to the PING nomenclature. For each cell, 38 features were extracted from responses to suprathreshold current stimuli and statistically analyzed to examine whether cortical interneurons subdivide into e-types. We showed that the partitioning into different e-types is indeed the major component of data variability. The analysis suggests refining the PING e-type classification to be hierarchical, whereby most variability is first captured within a coarse subpartition, and then subsequently divided into finer subpartitions. The coarse partition matches the well-known partitioning of interneurons into fast spiking and adapting cells. Finer subpartitions match the burst, continuous, and delayed subtypes. Additionally, our analysis enabled the ranking of features according to their ability to differentiate among e-types. We showed that our quantitative e-type assignment is more than 90% accurate and manages to catch several human errors.

Postnatal maturation of somatostatin-expressing inhibitory cells in the somatosensory cortex of GIN mice

Postnatal inhibitory neuron development affects mammalian brain function, and failure of this maturation process may underlie pathological conditions such as epilepsy, schizophrenia, and depression. Furthermore, understanding how physiological properties of inhibitory neurons change throughout development is critical to understanding the role(s) these cells play in cortical processing. One subset of inhibitory neurons that may be affected during postnatal development is somatostatin-expressing (SOM) cells. A subset of these cells is labeled with green-fluorescent protein (GFP) in a line of mice known as the GFP-positive inhibitory neurons (GIN) line. Here, we studied how intrinsic electrophysiological properties of these cells changed in the somatosensory cortex of GIN mice between postnatal ages P11 and P32+. GIN cells were targeted for whole-cell current-clamp recordings and ranges of positive and negative current steps were presented to each cell. The results showed that as the neocortical circuitry matured during this critical time period multiple intrinsic and firing properties of GIN inhibitory neurons, as well as those of excitatory (regular-spiking [RS]) cells, were altered. Furthermore, these changes were such that the output of GIN cells, but not RS cells, increased over this developmental period. We quantified changes in excitability by examining the input–output relationship of both GIN and RS cells. We found that the firing frequency of GIN cells increased with age, while the rheobase current remained constant across development. This created a multiplicative increase in the input–output relationship of the GIN cells, leading to increases in gain with age. The input–output relationship of the RS cells, on the other hand, showed primarily a subtractive shift with age, but no substantial change in gain. These results suggest that as the neocortex matures, inhibition coming from GIN cells may become more influential in the circuit and play a greater role in the modulation of neocortical activity.

Differentiation and functional incorporation of embryonic stem cell-derived GABAergic interneurons in the dentate gyrus of mice with temporal lobe epilepsy

Cell therapies for neurological disorders require an extensive knowledge of disease-associated neuropathology and procedures for generating neurons for transplantation. In many patients with severe acquired temporal lobe epilepsy (TLE), the dentate gyrus exhibits sclerosis and GABAergic interneuron degeneration. Mounting evidence suggests that therapeutic benefits can be obtained by transplanting fetal GABAergic progenitors into the dentate gyrus in rodents with TLE, but the scarcity of human fetal cells limits applicability in patient populations. In contrast, virtually limitless quantities of neural progenitors can be obtained from embryonic stem (ES) cells. ES cell-based therapies for neurological repair in TLE require evidence that the transplanted neurons integrate functionally and replace cell types that degenerate. To address these issues, we transplanted mouse ES cell-derived neural progenitors (ESNPs) with ventral forebrain identities into the hilus of the dentate gyrus of mice with TLE and evaluated graft differentiation, mossy fiber sprouting, cellular morphology, and electrophysiological properties of the transplanted neurons. In addition, we compared electrophysiological properties of the transplanted neurons with endogenous hilar interneurons in mice without TLE. The majority of transplanted ESNPs differentiated into GABAergic interneuron subtypes expressing calcium-binding proteins parvalbumin, calbindin, or calretinin. Global suppression of mossy fiber sprouting was not observed; however, ESNP-derived neurons formed dense axonal arborizations in the inner molecular layer and throughout the hilus. Whole-cell hippocampal slice electrophysiological recordings and morphological analyses of the transplanted neurons identified five basic types; most with strong after-hyperpolarizations and smooth or sparsely spiny dendritic morphologies resembling endogenous hippocampal interneurons. Moreover, intracellular recordings of spontaneous EPSCs indicated that the new cells functionally integrate into epileptic hippocampal circuitry.