<i>Intellicount</i>: High-Throughput Quantification of Fluorescent Synaptic Protein Puncta by Machine Learning

Fantuzzo, Joseph A.; Mirabella, Vincent R.; Hamod, Aula; Hart, Ronald P.; Zahn, Jeffrey D.; Pang, Zhiping P.

doi:10.1523/eneuro.0219-17.2017

Cited by 37 publications

(34 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ACSAT has three sets of free parameters that can be objectively chosen or are not sensitive: should be chosen based on how large neurons are expected to be using information including image resolution, magnification, etc. The presence of these area criteria in our algorithm is consistent with the literature (Fantuzzo et al, 2017).…”

Section: Discussionsupporting

confidence: 89%

“…Otsu's method is limited when identifying ROIs with different pixel intensities due to the uneven lighting of the background. A recent machine learning-based algorithm uses image gradients and pixel traces to automatically optimize the threshold value (Fantuzzo et al, 2017). However, the method still requires a user's subjective input in selecting a background removal factor based on the dataset.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Cell Segmentation by Adaptive Thresholding (ACSAT) for large scale calcium imaging datasets

Shen

Tseng

Hansen

et al. 2018

Preprint

View full text Add to dashboard Cite

Advances in calcium imaging have made it possible to record from an increasingly larger number of neurons simultaneously. Neuroscientists can now routinely image hundreds to thousands of individual neurons. With the continued neurotechnology development effort, it is expected that millions of neurons could soon be simultaneously measured. An emerging technical challenge that parallels the advancement in imaging such a large number of individual neurons is the processing of correspondingly large datasets, an important step of which is the identification of individual neurons. Traditional methods rely mainly on manual or semi-manual inspection, which cannot be scaled to processing large datasets. To address this challenge, we have developed an automated cell segmentation method, which is referred to as Automated Cell Segmentation by Adaptive Thresholding (ACSAT). ACSAT includes an iterative procedure that automatically calculates global and local threshold values during each iteration based on image pixel intensities. As such, the algorithm is capable of handling morphological variations and dynamic changes in fluorescence intensities in different calcium imaging datasets. In addition, ACSAT computes adaptive threshold values based on a time-collapsed image that is representative of the image sequence, and thus ACSAT provides segmentation results at a fast speed. We tested the algorithm on wide-field calcium imaging datasets in the brain regions of hippocampus and striatum in mice. ACSAT achieved precision and recall rates of approximately 80% when compared to individual neurons that are verified by human inspection. Additionally, ACSAT successfully detected low-intensity neurons that were initially undetected by humans.SignificanceACSAT automatically segments cells in large scale wide-field calcium imaging datasets. It is based on adaptive thresholding at both global and local levels, implemented in an iterative process to identify individual neurons in a time-collapsed image from an image sequence. It is therefore capable of handling variation in cell morphology and dynamic changes between different calcium imaging datasets at a fast speed. Based on tests performed on two datasets from mouse hippocampus and striatum, ACSAT performed comparable to human referees and was even able to detect low-intensity neurons that were initially undetected by human referees.

show abstract

Section: Discussionsupporting

confidence: 89%

Section: Introductionmentioning

confidence: 99%

Automatic Cell Segmentation by Adaptive Thresholding (ACSAT) for large scale calcium imaging datasets

Shen

Tseng

Hansen

et al. 2018

Preprint

View full text Add to dashboard Cite

show abstract

“…Total dendritic length was obtained by tracing 773 MAP2 expression using the NeuronJ plugin (Meijering et al, 2004). VGLUT puncta number was 774 calculated using Intellicount software (Fantuzzo et al, 2017). hypertonic sucrose solution directly onto the neuron and then divided the sucrose charge by the charge 796 of the average miniature event onto the same neuron (Rosenmund & Stevens, 1996).…”

Section: Immunocytochemistry 743mentioning

confidence: 99%

mTORC1 and mTORC2 regulate distinct aspects of glutamatergic synaptic transmission

McCabe

Cullen

Barrows

et al. 2019

Preprint

View full text Add to dashboard Cite

25Although mTOR signaling is known as a broad regulator of cell growth and proliferation, in 26 neurons it regulates synaptic transmission, which is thought to be a major mechanism through which 27 altered mTOR signaling leads to neurological disease. Although previous studies have delineated 28 postsynaptic roles for mTOR, whether it regulates presynaptic function is largely unknown. Moreover, 29 the mTOR kinase operates in two complexes, mTORC1 and mTORC2, suggesting that mTOR's role in 30 synaptic transmission may be complex-specific. To better understand each complex's role in synaptic 31 transmission, we genetically inactivated mTORC1 or mTORC2 in cultured mouse glutamatergic 32 hippocampal neurons. Inactivation of either complex reduced neuron growth and evoked EPSCs, 33 however, mTORC1 exerted its effects on eEPSCs at the postsynapse and mTORC2 at the presynapse. 34 Furthermore, inactivation of each complex altered specific modes of synaptic vesicle release, suggesting 35 that mTORC1 and mTORC2 differentially modulate postsynaptic responsiveness and presynaptic release 36 to optimize glutamatergic synaptic transmission. 37 38 39 40 41 42 43 44 45 103 Results 104 Inactivation of mTORC1 or mTORC2 in neurons, via Raptor or Rictor deletion, causes distinct effects on 105 downstream mTOR signaling. 106To investigate the effects of mTORC1 or mTORC2 inactivation on neurons, we cultured primary 107 hippocampal neurons isolated from P0-P1 Raptor flox/flox or Rictor flox/flox mice on astrocytes isolated from 108 P0-P1 wild-type mice. At the time of plating, we transduced the neurons with adeno-associated viruses 109 (AAVs) expressing either an mCherry-Cre fusion protein or mCherry alone, both driven by the hSyn 110 promoter, to generate knockout (Raptor-KO or Rictor-KO) or control (Raptor-Con or Rictor-Con) 111 neurons, respectively. To assess the efficiency of gene deletion in Raptor-KO and Rictor-KO neurons, we 112 performed quantitative immunofluorescence analysis for phospho-S6 (S240/244) (pS6), an indicator of 113 mTORC1 activity, and phospho-Akt (S473) (pAkt), an indicator of mTORC2 activity. In Raptor-KO neurons, 114 the mean fluorescence intensity of the pS6 signal was decreased relative to Raptor-Con neurons (Con: 1151.00 ± 0.05, Rap-KO: 0.25 ± 0.03, p < 0.001; Figure 1A1-3), verifying a reduction in mTORC1 activity. The 116 pAkt signal was, however, increased in Raptor-KO neurons relative to Raptor-Con neurons (Con: 1.00 ± 117

show abstract

“…Importantly, SACT is applicable to a variety of synaptic antigens with very different distributions, because the user defines the expected molecular composition and size of synapses where the antigen is present. Furthermore, the algorithm can be applied to new datasets without creating extensive manual annotations for each synapse subtype, unlike traditional classifiers such as support vector machines and deep learning used by other synapse detection algorithms (Bass et al, 2017;Busse and Smith, 2013;Collman et al, 2015;Fantuzzo et al, 2017;Kreshuk et al, 2014).…”

Section: Overviewmentioning

confidence: 99%

“…Characterizing synaptic antibodies for AT immunolabeling of brain sections requires detecting synapses. Over the past few years, several synapse detection methods have been presented which use traditional machine learning paradigms for detection (Bass et al, 2017;Busse and Smith, 2013;Collman et al, 2015;Fantuzzo et al, 2017;Kreshuk et al, 2014). While they perform well, each requires the user to The first box shows the raw image data from immunolabeling with one antibody.…”

Section: Synapse Detectionmentioning

confidence: 99%

A Computational Synaptic Antibody Characterization and Screening Framework for Array Tomography

Simhal

Gong

et al. 2018

Preprint

View full text Add to dashboard Cite

16Application-specific validation of antibodies is a critical prerequisite for their successful use. Here 17 we introduce an automated approach for characterization and screening of antibodies against 18 synaptic molecules for high-resolution immunofluorescence array tomography. The method, based 19 on a probabilistic synapse detection algorithm, is designed to provide a robust, flexible, and 20 efficient tool for antibody characterization. By allowing the user to define the molecular 21 composition and size of synapses expected to contain the antigen, the method is applicable to 22 synaptic molecules with different distribution. The output of the algorithm are automatically 23 computed characteristics such as synapse density and target specificity ratio, which reflect the 24 sensitivity and specificity of immunolabeling with a given antibody. These measurements provide 25 an objective way to characterize and compare the performance of different antibodies against the 26 same target, and can be used to objectively select the antibodies best suited for AT, and potentially 27 also for other immunolabeling applications. 28 29 30Antibodies are an indispensable tool for the modern biologist. Their high-affinity binding to spe-31 cific target molecules makes it possible to detect, isolate, and manipulate the function of these 32 molecules. A staggering number of antibodies are available to the research community, as are 33 many options to make new antibodies. However, since antibodies are biological tools employed in 34 complex systems, they can be very difficult to evaluate and to use in a predictable and reproducible 35 way. Indeed, a large volume of misleading or incorrect data has been published based on results 36 from antibodies that did not perform as assumed (Anderson and Grant, 2006; Baker, 2015; Rhodes 37 and Trimmer, 2006). 38 1 of 21 Preprint for biorxiv.org 39 Recognizing this problem, there has been substantial progress in optimizing antibody produc-40 tion (Nilsson et al., 2005; Gong et al., 2016). The importance of establishing reliable practices for 41 antibody use is now widely recognized, and many companies are adopting transparent practices 42 for rigorous antibody validation (Fritschy et al., 1998; Uhlen et al., 2016). The performance of anti-43 bodies, however, is application-specific (Lorincz and Nusser, 2008), and the reliable performance 44 of an antibody in one application does not guarantee its suitability for another application. For 45 example, an antibody that yields a single band on an immunoblot analysis of a tissue homogenate 46 may prove wholly unsatisfactory for immunohistochemistry on sections of fixed tissue. Moreover, 47 the same antibody that yields a robust and specific signal in immunohistochemical labeling of 48 tissue sections prepared under one set of conditions may yield little or no signal on comparable 49 samples prepared under different conditions (Fritschy et al., 1998; Fukaya and Watanabe, 2000). 50 For practical reasons, commercial antibodie...

show abstract

Intellicount: High-Throughput Quantification of Fluorescent Synaptic Protein Puncta by Machine Learning

Cited by 37 publications

References 31 publications

Automatic Cell Segmentation by Adaptive Thresholding (ACSAT) for large scale calcium imaging datasets

Automatic Cell Segmentation by Adaptive Thresholding (ACSAT) for large scale calcium imaging datasets

mTORC1 and mTORC2 regulate distinct aspects of glutamatergic synaptic transmission

A Computational Synaptic Antibody Characterization and Screening Framework for Array Tomography

Contact Info

Product

Resources

About