Fast clustering and cell-type annotation of scATAC data using pre-trained embeddings.

Data from the single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) are now widely available. One major computational challenge is dealing with high dimensionality and inherent sparsity, which is typically addressed by producing lower dimensional representations of single cells for downstream clustering tasks. Current approaches produce such individual cell embeddings directly through a one-step learning process. Here, we propose an alternative approach by building embedding models pre-trained on reference data. We argue that this provides a more flexible analysis workflow that also has computational performance advantages through transfer learning. We implemented our approach in scEmbed, an unsupervised machine-learning framework that learns low-dimensional embeddings of genomic regulatory regions to represent and analyze scATAC-seq data. scEmbed performs well in terms of clustering ability and has the key advantage of learning patterns of region co-occurrence that can be transferred to other, unseen datasets. Moreover, models pre-trained on reference data can be exploited to build fast and accurate cell-type annotation systems without the need for other data modalities. scEmbed is implemented in Python and it is available to download from GitHub. We also make our pre-trained models available on huggingface for public use. scEmbed is open source and available at https://github.com/databio/geniml. Pre-trained models from this work can be obtained on huggingface: https://huggingface.co/databio.

The role of working memory capacity in the temporal compression of episodic memories: An individual differences approach.

Remembering past events usually takes less time than their actual duration-their unfolding is temporally compressed in episodic memory. The rate of temporal compression (i.e., the ratio of the actual duration of an event to the duration of its remembering) is not constant but varies between individuals and as a function of the structure of events (e.g., how they can be divided into shorter subevents). However, the cognitive mechanisms underlying these variations remain poorly understood. Given its role in the encoding and retrieval of information in episodic memory, working memory (WM) capacity could be an important determinant of temporal compression rates. We tested this hypothesis in two experiments in which we asked participants to watch and then mentally replay short videos showing people engaged in daily life activities. We showed that temporal compression rates depend on an interplay between WM and the structure of the remembered events: participants' WM capacity (assessed using complex span tasks) was negatively associated with temporal compression rates, but only when the remembered events contained few event boundaries (i.e., few subevents). This suggests that the temporal compression of events in episodic memory emerges when some of the subevents to be retained are too long to be fully represented in WM. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Working memory capacity for continuous events: The root of temporal compression in episodic memory?

Remembering the unfolding of past episodes usually takes less time than their actual duration. In this study, we evaluated whether such temporal compression emerges when continuous events are too long to be fully held in working memory. To do so, we asked 90 young adults to watch and mentally replay video clips showing people performing a continuous action (e.g., turning a car jack) that lasted 3, 6, 9, 12, or 15 s. For each clip, participants had to carefully watch the event and then to mentally replay it as accurately and precisely as possible. Results showed that mental replay durations increased with event duration but in a non-linear manner: they were close to the actual event duration for short videos (3-9 s), but significantly smaller for longer videos (12 and 15 s). These results suggest that working memory is temporally limited in its capacity to represent continuous events, which could in part explain why the unfolding of events is temporally compressed in episodic memory.

Joint Representation Learning for Retrieval and Annotation of Genomic Interval Sets.

As available genomic interval data increase in scale, we require fast systems to search them. A common approach is simple string matching to compare a search term to metadata, but this is limited by incomplete or inaccurate annotations. An alternative is to compare data directly through genomic region overlap analysis, but this approach leads to challenges like sparsity, high dimensionality, and computational expense. We require novel methods to quickly and flexibly query large, messy genomic interval databases. Here, we develop a genomic interval search system using representation learning. We train numerical embeddings for a collection of region sets simultaneously with their metadata labels, capturing similarity between region sets and their metadata in a low-dimensional space. Using these learned co-embeddings, we develop a system that solves three related information retrieval tasks using embedding distance computations: retrieving region sets related to a user query string, suggesting new labels for database region sets, and retrieving database region sets similar to a query region set. We evaluate these use cases and show that jointly learned representations of region sets and metadata are a promising approach for fast, flexible, and accurate genomic region information retrieval.

PEPhub: a database, web interface, and API for editing, sharing, and validating biological sample metadata.

Background: As biological data increases, we need additional infrastructure to share it and promote interoperability. While major effort has been put into sharing data, relatively less emphasis is placed on sharing metadata. Yet, sharing metadata is also important, and in some ways has a wider scope than sharing data itself. Results: Here, we present PEPhub, an approach to improve sharing and interoperability of biological metadata. PEPhub provides an API, natural language search, and user-friendly web-based sharing and editing of sample metadata tables. We used PEPhub to process more than 100,000 published biological research projects and index them with fast semantic natural language search. PEPhub thus provides a fast and user-friendly way to finding existing biological research data, or to share new data. Availability: https://pephub.databio.org.

GEOfetch: a command-line tool for downloading data and standardized metadata from GEO and SRA.

MOTIVATION: The Gene Expression Omnibus has become an important source of biological data for secondary analysis. However, there is no simple, programmatic way to download data and metadata from Gene Expression Omnibus (GEO) in a standardized annotation format. RESULTS: To address this, we present GEOfetch-a command-line tool that downloads and organizes data and metadata from GEO and SRA. GEOfetch formats the downloaded metadata as a Portable Encapsulated Project, providing universal format for the reanalysis of public data. AVAILABILITY AND IMPLEMENTATION: GEOfetch is available on Bioconda and the Python Package Index (PyPI).

GenomicDistributions: fast analysis of genomic intervals with Bioconductor.

BACKGROUND: Epigenome analysis relies on defined sets of genomic regions output by widely used assays such as ChIP-seq and ATAC-seq. Statistical analysis and visualization of genomic region sets is essential to answer biological questions in gene regulation. As the epigenomics community continues generating data, there will be an increasing need for software tools that can efficiently deal with more abundant and larger genomic region sets. Here, we introduce GenomicDistributions, an R package for fast and easy summarization and visualization of genomic region data. RESULTS: GenomicDistributions offers a broad selection of functions to calculate properties of genomic region sets, such as feature distances, genomic partition overlaps, and more. GenomicDistributions functions are meticulously optimized for best-in-class speed and generally outperform comparable functions in existing R packages. GenomicDistributions also offers plotting functions that produce editable ggplot objects. All GenomicDistributions functions follow a uniform naming scheme and can handle either single or multiple region set inputs. CONCLUSIONS: GenomicDistributions offers a fast and scalable tool for exploratory genomic region set analysis and visualization. GenomicDistributions excels in user-friendliness, flexibility of outputs, breadth of functions, and computational performance. GenomicDistributions is available from Bioconductor ( https://bioconductor.org/packages/release/bioc/html/GenomicDistributions.html ).

The Neural Representation of Ordinal Information: Domain-Specific or Domain-General?

Ordinal processing allows for the representation of the sequential relations between stimuli and is a fundamental aspect of different cognitive domains such as verbal working memory (WM), language and numerical cognition. Several studies suggest common ordinal coding mechanisms across these different domains but direct between-domain comparisons of ordinal coding are rare and have led to contradictory evidence. This fMRI study examined the commonality of ordinal representations across the WM, the number, and the letter domains by using a multivoxel pattern analysis approach and by focusing on triplet stimuli associated with robust ordinal distance effects. Neural patterns in fronto-parietal cortices distinguished ordinal distance in all domains. Critically, between-task predictions of ordinal distance in fronto-parietal cortices were robust between serial order WM, alphabetical order judgment but not when involving the numerical order judgment tasks. Moreover, frontal ROIs further supported between-task prediction of distance for the luminance judgment control task, the serial order WM, and the alphabetical tasks. These results suggest that common neural substrates characterize processing of ordinal information in WM and alphabetical but not numerical domains. This commonality, particularly in frontal cortices, may however reflect attentional control processes involved in judging ordinal distances rather than the intervention of domain-general ordinal codes.

Neuron Activity Dependent Redox Compartmentation Revealed with a Second Generation Red-Shifted Ratiometric Sensor.

Oxidative stress is a hallmark of several aging and trauma related neurological disorders, but the precise details of how altered neuronal activity elicits subcellular redox changes have remained difficult to resolve. Current redox sensitive dyes and fluorescent proteins can quantify spatially distinct changes in reactive oxygen species levels, but multicolor probes are needed to accurately analyze compartment-specific redox dynamics in single cells that can be masked by population averaging. We previously engineered genetically encoded red-shifted redox-sensitive fluorescent protein sensors using a Förster resonance energy transfer relay strategy. Here, we developed a second-generation excitation ratiometric sensor called rogRFP2 with improved red emission for quantitative live-cell imaging. Using this sensor to measure activity-dependent redox changes in individual cultured neurons, we observed an anticorrelation in which mitochondrial oxidation was accompanied by a concurrent reduction in the cytosol. This behavior was dependent on the activity of Complex I of the mitochondrial electron transport chain and could be modulated by the presence of cocultured astrocytes. We also demonstrated that the red fluorescent rogRFP2 facilitates ratiometric one- and two-photon redox imaging in rat brain slices and Drosophila retinas. Overall, the proof-of-concept studies reported here demonstrate that this new rogRFP2 redox sensor can be a powerful tool for understanding redox biology both in vitro and in vivo across model organisms.

The temporal compression of events during episodic future thinking.

While the cognitive and neural mechanisms that underlie episodic future thinking are increasingly well understood, little is known about how the temporal unfolding of events is represented in future simulations. In this study, we leveraged wearable camera technology to examine whether real-world events are structured and compressed in the same way when imagining the future as when remembering the past. We found that future events were simulated at proportionally higher speed than past events and that the density of experience units representing the unfolding of events was lower for future than for past episodes. Despite these differences, the nature of events influenced compression rates in the same way for past and future events. Furthermore, the perceived duration of both types of events depended on the density of represented experience units. These results provide novel insight into the mechanisms that structure the unfolding of events during future simulations.