Attention-based CNN-BiLSTM for sleep state classification of spatiotemporal wide-field calcium imaging data.

BACKGROUND: Wide-field calcium imaging (WFCI) with genetically encoded calcium indicators allows for spatiotemporal recordings of neuronal activity in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming, invasive and often suffers from low inter- and intra-rater reliability. Therefore, an automated sleep state classification method that operates on spatiotemporal WFCI data is desired. NEW METHOD: A hybrid network architecture consisting of a convolutional neural network (CNN) to extract spatial features of image frames and a bidirectional long short-term memory network (BiLSTM) with attention mechanism to identify temporal dependencies among different time points was proposed to classify WFCI data into states of wakefulness, NREM and REM sleep. RESULTS: Sleep states were classified with an accuracy of 84 % and Cohen's κ of 0.64. Gradient-weighted class activation maps revealed that the frontal region of the cortex carries more importance when classifying WFCI data into NREM sleep while posterior area contributes most to the identification of wakefulness. The attention scores indicated that the proposed network focuses on short- and long-range temporal dependency in a state-specific manner. COMPARISON WITH EXISTING METHOD: On a held out, repeated 3-hour WFCI recording, the CNN-BiLSTM achieved a κ of 0.67, comparable to a κ of 0.65 corresponding to the human EEG/EMG-based scoring. CONCLUSIONS: The CNN-BiLSTM effectively classifies sleep states from spatiotemporal WFCI data and will enable broader application of WFCI in sleep research.

Validation of Deep Learning-based Sleep State Classification.

Deep learning methods have been developed to classify sleep states of mouse electroencephalogram (EEG) and electromyogram (EMG) recordings with accuracy reported as high as 97%. However, when applied to independent datasets, with a variety of experimental and recording conditions, sleep state classification accuracy often drops due to distributional shift. Mixture z-scoring, a pre-processing standardization of EEG/EMG signals, has been suggested to account for these variations. This study sought to validate mixture z-scoring in combination with a deep learning method on an independent dataset. The open-source software Accusleep, which implements mixture z-scoring in combination with deep learning via a convolutional neural network, was used to classify sleep states in 12, three-hour EEG/EMG recordings from mice sleeping in a head-fixed position. Mixture z-scoring with deep learning classified sleep states on two independent recordings with 85-92% accuracy and a Cohen's κ of 0.66-0.71. These results validate mixture z-scoring in combination with deep learning to classify sleep states with the potential for widespread use.

Normal aging in mice is associated with a global reduction in cortical spectral power and network-specific declines in functional connectivity.

Normal aging is associated with a variety of neurologic changes including declines in cognition, memory, and motor activity. These declines correlate with neuronal changes in synaptic structure and function. Degradation of brain network activity and connectivity represents a likely mediator of age-related functional deterioration resulting from these neuronal changes. Human studies have demonstrated both general decreases in spontaneous cortical activity and disruption of cortical networks with aging. Current techniques used to study cerebral network activity are hampered either by limited spatial resolution (e.g. electroencephalography, EEG) or limited temporal resolution (e.g., functional magnetic resonance imaging, fMRI). Here we utilize mesoscale imaging of neuronal activity in Thy1-GCaMP6f mice to characterize neuronal network changes in aging with high spatial resolution across a wide frequency range. We show that while evoked activity is unchanged with aging, spontaneous neuronal activity decreases across a wide frequency range (0.01-4 Hz) involving all regions of the cortex. In contrast to this global reduction in cortical power, we found that aging is associated with functional connectivity (FC) deterioration of select networks including somatomotor, cingulate, and retrosplenial nodes. These changes are corroborated by reductions in homotopic FC and node degree within somatomotor and visual cortices. Finally, we found that whole-cortex delta power and delta band node degree correlate with exploratory activity in young but not aged animals. Together these data suggest that aging is associated with global declines in spontaneous cortical activity and focal deterioration of network connectivity, and that these reductions may be associated with age-related behavioral declines.

Alternative expression analysis by RNA sequencing.

In alternative expression analysis by sequencing (ALEXA-seq), we developed a method to analyze massively parallel RNA sequence data to catalog transcripts and assess differential and alternative expression of known and predicted mRNA isoforms in cells and tissues. As proof of principle, we used the approach to compare fluorouracil-resistant and -nonresistant human colorectal cancer cell lines. We assessed the sensitivity and specificity of the approach by comparison to exon tiling and splicing microarrays and validated the results with reverse transcription-PCR, quantitative PCR and Sanger sequencing. We observed global disruption of splicing in fluorouracil-resistant cells characterized by expression of new mRNA isoforms resulting from exon skipping, alternative splice site usage and intron retention. Alternative expression annotation databases, source code, a data viewer and other resources to facilitate analysis are available at http://www.alexaplatform.org/alexa_seq/.

SPARC in cancer biology: its role in cancer progression and potential for therapy.

The ability to effectively target a tumor to achieve complete regression and cure is the ultimate goal that drives our need to better understand tumor biology. Recently, SPARC has generated considerable interest as a multi-faceted protein that belongs to a family of matricellular proteins. It functions not only to modulate cell-cell and cell-matrix interactions, but its de-adhesive and growth inhibitory properties in non-transformed cells have led to studies to assess its role in cancer. Its divergent actions reflect the complexity of this protein, because in certain types of cancers, such as melanomas and gliomas, SPARC is associated with a highly aggressive tumor phenotype, while in others, mainly ovarian, neuroblastomas and colorectal cancers, SPARC may function as a tumor suppressor. Recent studies have also demonstrated a role for SPARC in sensitizing therapy-resistant cancers. Here, the role of SPARC in cancer progression and its potential application in cancer therapy is discussed.

A novel interaction between procaspase 8 and SPARC enhances apoptosis and potentiates chemotherapy sensitivity in colorectal cancers.

Chemotherapy resistance accounts for the high mortality rates in patients with advanced cancers. We previously used a genomics approach to determine novel genes associated with this phenomenon and identified secreted protein acidic and rich in cysteine (SPARC) as a chemosensitizer capable of reversing therapy resistance in colorectal cancer cells by enhancing apoptosis in vitro and tumor regression in vivo. Here, we examined the mechanisms by which SPARC enhances apoptosis in the presence of chemotherapy. We show that SPARC potentiates apoptosis by augmenting the signaling cascade in a caspase-8-dependent manner, because apoptosis can be abolished by caspase 8 small interfering RNA in the presence of SPARC. This occurs independently of death receptor activation and leads to downstream involvement of Bid and subsequent apoptosis. Interestingly, this results from an interaction between SPARC and the N terminus of the procaspase-8 DED-containing domain. These exciting findings provide an initial map of the apoptosis signaling events mediated by SPARC and how this can ultimately result in the reversal of chemotherapy resistance and enhanced tumor regression. This signaling cascade can be exploited therapeutically and may have potential clinical implications for patients with advanced and therapy-refractory cancers.

Synergism between vitamin D and secreted protein acidic and rich in cysteine-induced apoptosis and growth inhibition results in increased susceptibility of therapy-resistant colorectal cancer cells to chemotherapy.

Patients with advanced colorectal cancer continue to have poor outcomes because of therapy-refractory disease. We previously showed that secreted protein acidic and rich in cysteine (SPARC) gene and protein could function as a chemotherapy sensitizer by enhancing tumor regression in response to radiation and chemotherapy in tumor xenograft models of chemotherapy-resistant tumors. This function of SPARC was gleamed from a microarray analysis that also revealed down-regulation of the vitamin D receptor (VDR) in therapy-refractory colorectal cancer cells. This study examines the potential synergistic effect of SPARC and vitamin D, which up-regulates VDR, in enhancing chemotherapy response in colorectal cancer. Using MIP101 colorectal cancer cell lines and SPARC-overexpressing MIP101 cells, we were able to show that, in the presence of SPARC, exposure to low doses of 1alpha,25-dihydroxyvitamin D(3) significantly reduces cell viability, enhances chemotherapy-induced apoptosis, and inhibits the growth of colorectal cancer cells. Moreover, in tumor xenograft mouse models, up-regulation of VDR was seen in tumors that had the greatest regression following treatment that combined SPARC with chemotherapy. Therefore, our findings reveal a synergistic effect between SPARC and low doses of 1alpha,25-dihydroxyvitamin D(3) that further augments the sensitivity of tumors to chemotherapy. This combination may prove to be a useful adjunct in the treatment of colorectal cancer, especially in those patients with therapy-refractory disease.