Target speech separation refers to extracting a target speaker's voice from an overlapped audio of simultaneous talkers. Previously the use of visual modality for target speech separation has demonstrated great potentials. This work proposes a general multi-modal framework for target speech separation by utilizing all the available information of the target speaker, including his/her spatial location, voice characteristics and lip movements. Also, under this framework, we investigate on the fusion methods for multi-modal joint modeling. A factorized attention-based fusion method is proposed to aggregate the high-level semantic information of multi-modalities at embedding level. This method firstly factorizes the mixture audio into a set of acoustic subspaces, then leverages the target's information from other modalities to enhance these subspace acoustic embeddings with a learnable attention scheme. To validate the robustness of proposed multi-modal separation model in practical scenarios, the system was evaluated under the condition that one of the modalities is temporarily missing, invalid or corrupted. Experiments are conducted on a large-scale audio-visual dataset collected from YouTube (to be released) that spatialized by simulated room impulse responses (RIRs). Experiment results illustrate that our proposed multi-modal framework significantly outperforms single-modal and bi-modal speech separation approaches, while can still support real-time processing.

In recent years, the advances in single-cell RNA-seq techniques have enabled us to perform large-scale transcriptomic profiling at single-cell resolution in a high-throughput manner. Unsupervised learning such as data clustering has become the central component to identify and characterize novel cell types and gene expression patterns. In this study, we review the existing single-cell RNA-seq data clustering methods with critical insights into the related advantages and limitations. In addition, we also review the upstream single-cell RNA-seq data processing techniques such as quality control, normalization, and dimension reduction. We conduct performance comparison experiments to evaluate several popular single-cell RNA-seq clustering approaches on simulated and multiple single-cell transcriptomic data sets.

The off‐target effects induced by guide RNAs in the CRISPR/Cas9 gene‐editing system have raised substantial concerns in recent years. Many in silico predictive models have been developed for predicting the off‐target activities; however, few are capable of predicting the off‐target activities with insertions or deletions between guide RNA and target DNA sequence pair. In order to fill this gap, a recurrent convolutional network named CRISPR‐Net is developed for scoring the gRNA‐target pairs with mismatches and indels; and a machine‐learning based model named CRISPR‐Net‐Aggregate is also developed for aggregating the scores as the consensus off‐target score for each potential guide RNA. It is demonstrated that CRISPR‐Net achieves competitive performance on CIRCLE‐Seq and GUIDE‐seq datasets with indels and mismatches, outperforming the state‐of‐the‐art off‐target prediction methods on two independent mismatch‐only datasets. The CRISPR‐Net‐Aggregate also surpasses a competing method on the aggregation task. Moreover, a two‐stage sensitivity analysis is introduced to visualize the CRISPR‐Net prediction on the gRNA‐target pair of interest, demonstrating how implicit knowledge encoded in CRISPR‐Net contributes to the accurate off‐target activity quantification. Finally, the source code is made available at the Code Ocean repository (https://codeocean.com/capsule/9553651/tree/v1).

Background noise, interfering speech and room reverberation frequently distort target speech in real listening environments. In this study, we address joint speech separation and dereverberation, which aims to separate target speech from background noise, interfering speech and room reverberation. In order to tackle this fundamentally difficult problem, we propose a novel multimodal network that exploits both audio and visual signals. The proposed network architecture adopts a two-stage strategy, where a separation module is employed to attenuate background noise and interfering speech in the first stage and a dereverberation module to suppress room reverberation in the second stage. The two modules are first trained separately, and then integrated for joint training, which is based on a new multi-objective loss function. Our experimental results show that the proposed multimodal network yields consistently better objective intelligibility and perceptual quality than several one-stage and two-stage baselines. We find that our network achieves a 21.10% improvement in ESTOI and a 0.79 improvement in PESQ over the unprocessed mixtures. Moreover, our network architecture does not require the knowledge of the number of speakers.

The RNA-guided CRISPR/Cas9 system has been widely applied to genome editing. CRISPR/Cas9 system can effectively edit the on-target genes. Nonetheless, it has recently been demonstrated that many homologous off-target genomic sequences could be mutated, leading to unexpected gene-editing outcomes. Therefore, a plethora of tools were proposed for the prediction of off-target activities of CRISPR/Cas9. Nonetheless, each computational tool has its own advantages and drawbacks under diverse conditions. It is hardly believed that a single tool is optimal for all conditions. Hence, we would like to explore the ensemble learning potential on synergizing multiple tools with genomic annotations together to enhance its predictive abilities. We proposed an ensemble learning framework which synergizes multiple tools together to predict the off-target activities of CRISPR/Cas9 in different combinations. Interestingly, the ensemble learning using AdaBoost outperformed other individual off-target predictive tools. We also investigated the effect of evolutionary conservation (PhyloP and PhastCons) and chromatin annotations (ChromHMM and Segway) and found that only PhyloP can enhance the predictive capabilities further. Case studies are conducted to reveal ensemble insights into the off-target predictions, demonstrating how the current study can be applied in different genomic contexts. The best prediction predicted by AdaBoost is up to 0.9383 (AUC) and 0.2998 (PRC) that outperforms other classifiers. This is ascribable to the fact that AdaBoost introduces a new weak classifier (i.e. decision stump) in each iteration to learn the DNA sequences that were misclassified as off-targets until a small error rate is reached iteratively. The source codes are freely available on GitHub at https://github.com/Alexzsx/CRISPR. Supplementary data are available at Bioinformatics online.

Many purely neural network based speech separation approaches have been proposed to improve objective assessment scores, but they often introduce nonlinear distortions that are harmful to modern automatic speech recognition (ASR) systems. Minimum variance distortionless response (MVDR) filters are often adopted to remove nonlinear distortions, however, conventional neural mask-based MVDR systems still result in relatively high levels of residual noise. Moreover, the matrix inverse involved in the MVDR solution is sometimes numerically unstable during joint training with neural networks. In this study, we propose a multi-channel multi-frame (MCMF) all deep learning (ADL)-MVDR approach for target speech separation, which extends our preliminary multi-channel ADL-MVDR approach. The proposed MCMF ADL-MVDR system addresses linear and nonlinear distortions. Spatio-temporal cross correlations are also fully utilized in the proposed approach. The proposed systems are evaluated using a Mandarin audio-visual corpus and are compared with several state-of-the-art approaches. Experimental results demonstrate the superiority of our proposed systems under different scenarios and across several objective evaluation metrics, including ASR performance.