Tradeoffs in alignment and assembly-based methods for structural variant detection with long-read sequencing data.

Long-read sequencing offers long contiguous DNA fragments, facilitating diploid genome assembly and structural variant (SV) detection. Efficient and robust algorithms for SV identification are crucial with increasing data availability. Alignment-based methods, favored for their computational efficiency and lower coverage requirements, are prominent. Alternative approaches, relying solely on available reads for de novo genome assembly and employing assembly-based tools for SV detection via comparison to a reference genome, demand significantly more computational resources. However, the lack of comprehensive benchmarking constrains our comprehension and hampers further algorithm development. Here we systematically compare 14 read alignment-based SV calling methods (including 4 deep learning-based methods and 1 hybrid method), and 4 assembly-based SV calling methods, alongside 4 upstream aligners and 7 assemblers. Assembly-based tools excel in detecting large SVs, especially insertions, and exhibit robustness to evaluation parameter changes and coverage fluctuations. Conversely, alignment-based tools demonstrate superior genotyping accuracy at low sequencing coverage (5-10×) and excel in detecting complex SVs, like translocations, inversions, and duplications. Our evaluation provides performance insights, highlighting the absence of a universally superior tool. We furnish guidelines across 31 criteria combinations, aiding users in selecting the most suitable tools for diverse scenarios and offering directions for further method development.

Neural Mechanisms of Working Memory Accuracy Revealed by Recurrent Neural Networks.

Understanding the neural mechanisms of working memory has been a long-standing Neuroscience goal. Bump attractor models have been used to simulate persistent activity generated in the prefrontal cortex during working memory tasks and to study the relationship between activity and behavior. How realistic the assumptions of these models are has been a matter of debate. Here, we relied on an alternative strategy to gain insights into the computational principles behind the generation of persistent activity and on whether current models capture some universal computational principles. We trained Recurrent Neural Networks (RNNs) to perform spatial working memory tasks and examined what aspects of RNN activity accounted for working memory performance. Furthermore, we compared activity in fully trained networks and immature networks, achieving only imperfect performance. We thus examined the relationship between the trial-to-trial variability of responses simulated by the network and different aspects of unit activity as a way of identifying the critical parameters of memory maintenance. Properties that spontaneously emerged in the artificial network strongly resembled persistent activity of prefrontal neurons. Most importantly, these included drift of network activity during the course of a trial that was causal to the behavior of the network. As a consequence, delay period firing rate and behavior were positively correlated, in strong analogy to experimental results from the prefrontal cortex. These findings reveal that delay period activity is computationally efficient in maintaining working memory, as evidenced by unbiased optimization of parameters in artificial neural networks, oblivious to the properties of prefrontal neurons.

Emergence of prefrontal neuron maturation properties by training recurrent neural networks in cognitive tasks.

Working memory and response inhibition are functions that mature relatively late in life, after adolescence, paralleling the maturation of the prefrontal cortex. The link between behavioral and neural maturation is not obvious, however, making it challenging to understand how neural activity underlies the maturation of cognitive function. To gain insights into the nature of observed changes in prefrontal activity between adolescence and adulthood, we investigated the progressive changes in unit activity of recurrent neural networks as they were trained to perform working memory and response inhibition tasks. These included increased delay period activity during working memory tasks and increased activation in antisaccade tasks. These findings reveal universal properties underlying the neuronal computations behind cognitive tasks and explicate the nature of changes that occur as the result of developmental maturation.

Aquila_stLFR: diploid genome assembly based structural variant calling package for stLFR linked-reads.

Motivation: Identifying structural variants (SVs) is critical in health and disease, however, detecting them remains a challenge. Several linked-read sequencing technologies, including 10X Genomics, TELL-Seq and single tube long fragment read (stLFR), have been recently developed as cost-effective approaches to reconstruct multi-megabase haplotypes (phase blocks) from sequence data of a single sample. These technologies provide an optimal sequencing platform to characterize SVs, though few computational algorithms can utilize them. Thus, we developed Aquila_stLFR, an approach that resolves SVs through haplotype-based assembly of stLFR linked-reads. Results: Aquila_stLFR first partitions long fragment reads into two haplotype-specific blocks with the assistance of the high-quality reference genome, by taking advantage of the potential phasing ability of the linked-read itself. Each haplotype is then assembled independently, to achieve a complete diploid assembly to finally reconstruct the genome-wide SVs. We benchmarked Aquila_stLFR on a well-studied sample, NA24385, and showed Aquila_stLFR can detect medium to large size deletions (50 bp-10 kb) with high sensitivity and medium-size insertions (50 bp-1 kb) with high specificity. Availability and implementation: Source code and documentation are available on https://github.com/maiziex/Aquila_stLFR. Supplementary information: Supplementary data are available at Bioinformatics Advances online.