Bloom.

□ Transition to hyperchaos and rare large-intensity pulses in Zeeman laser

>> https://pubs.aip.org/aip/cha/article/33/2/023128/2876208/Transition-to-hyperchaos-and-rare-large-intensity

Hyperchaos appears with a sudden expansion of the attractor of the system at a critical parameter for each case and it coincides with triggering of occasional and recurrent large-intensity pulses.

The transition to hyperchaos from a periodic orbit via Pomeau-Manneville intermittency shows hysteresis at the critical point, while no hysteresis is recorded during the other two processes.

Intriguingly, the transition to large-intensity pulses and the hyperchaotic dynamics appear concurrently, which is confirmed by the existence of two positive Lyapunov exponents in the system.





□ FlowShape: Cell shape characterization, alignment and comparison

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad383/7199619

FlowShape, a framework to describe cell shapes completely and to a tunable degree of detail. First, the procedure maps the mean curvature of the shape onto the sphere, resulting in a single function. This reduces the complexity associated with using multiple coordinate functions.

This function is decomposed into Spherical Harmonics to capture shape information. This Spherical Harmonics representation is then used to align, average and compare cell shapes, as well as to detect specific features, such as protrusions.





□ MultiVI: deep generative model for the integration of multimodal data

>> https://www.nature.com/articles/s41592-023-01909-9

MultiVI provides solutions for the two levels of analysis, with a low-dimensional summary of cell state and a normalized high-dimensional view of both modalities (measured or inferred) in each cell.

MultiVI was designed to account for the general caveats of single-cell genomics data, namely batch effects, variability in sequencing depth, limited sensitivity and noise. MultiVI integrates paired and single-modality data into a common low-dimensional representation.





□ MEvA-X: A Hybrid Multi-Objective Evolutionary Tool Using an XGBoost Classifier for Biomarkers Discovery on Biomedical Datasets

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad384/7199580

MEvA-X, a novel hybrid ensemble for feature selection and classification, combining a niche-based multi-objective evolutionary algorithm (EA) with the XGBoost classifier.

MEvA-X deploys a multi-objective EA to optimize the hyper-parameters of the classifier and perform feature selection, identifying a set of Pareto-optimal solutions and optimizing multiple objectives, including classification and model simplicity metrics.





□ DynamicViz: Dynamic visualization of high-dimensional data

>> https://www.nature.com/articles/s43588-022-00380-4

Dynamic visualizations can help to discriminate robust bridging connections that appear across most bootstrap visualizations from incidental or artificial bridging connections that only appear in one or a small minority of bootstrap visualizations.

Dynamic visualization with stacked integration of bootstrap visualizations generates static Portable Network Graphics. Stacked visualization overlays all bootstrap visualizations with user-defined opacity, offering orthogonal information to interactive or animated visualizations.





□ BGCFlow: Systematic pangenome workflow for the analysis of biosynthetic gene clusters across large genomic datasets

>> https://www.biorxiv.org/content/10.1101/2023.06.14.545018v1

BGCflow, a versatile Snakemake workflow aimed to aid large-scale genome mining studies to comprehensively analyze the secondary metabolite potential of selected bacterial species.

BGCflow integrates various genome analytics tools for organizing sample metadata, data selection, functional annotation, genome mining, phylogenetic placement, and comparative genomics.





□ MultiNicheNet: a flexible framework for differential cell-cell communication analysis from multi-sample multi-condition single-cell transcriptomics data

>> https://www.biorxiv.org/content/10.1101/2023.06.13.544751v1

MultiNicheNet builds upon the principles of SOTA for DE analysis. The algorithm considers inter-sample heterogeneity, can correct for batch effects and covariates, and can cope with complex experimental designs to address more challenging questions than pairwise comparisons.

MultiNicheNet uses this DE output to combine the principles of NicheNet and
ligand-receptor inference tools into one flexible framework. This enables the prioritization of ligand-receptor interactions based on DE, cell-type specific expression, and NicheNet's ligand activity.





□ BBmix: a Bayesian beta-binomial mixture model for accurate genotyping from RNA-sequencing

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad393/7203797

BBmix (Bayesian beta-binomial mixture model), a two-step method based on first modelling the genotype-specific read counts using beta-binomial distributions and then using these to infer genotype posterior probabilities.

BBmix can be incorporated into standard pipelines for calling genotypes. These parameters are generally transferable within datasets, such that a single learning run of less than one hour is sufficient to call genotypes in a large number of samples.





□ FiniMOM: Genetic fine-mapping from summary data using a non-local prior improves detection of multiple causal variants

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad396/7205323

FiniMOM (fine-mapping using a product inverse-moment prior), a novel Bayesian fine-mapping for summarized genetic associations. For causal effects, FiniMOM uses a non-local inverse-moment prior, which is a natural prior distribution to model non-null effects in finite samples.

A beta-binomial prior is set for the number of causal variants, with a parameterization that can be used to control for potential misspecifications in the linkage disequilibrium (LD) reference.





□ enviRule: An End-to-end System for Automatic Extraction of Reaction Patterns from Environmental Contaminant Biotransformation Pathways

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad407/7206883

enviRule, an automatic rule generation tool that can automatically extract rules from biotransformation, efficiently update automatic rules as new data is added, and determine the optimum genericity of rules for the task of contaminant pathway prediction using the enviPath.

enviRule consists of three modules, namely reaction clusterer, rule generator, and reaction adder, which work closely together to generate automatic rules. Reactions are fclustered in reaction clusterer based on reaction centers, then rule generator produces automatic rules.





□ RAD21 is the core subunit of the cohesin complex involved in directing genome organization

>> https://genomebiology.biomedcentral.com/articles/10.1186/s13059-023-02982-1

Directly visualizing the up-regulation of RAD21 leads to excessive chromatin loop extrusion into a vermicelli-like morphology with RAD21 clustered into foci and excessively loaded cohesin bow-tying a TAD to form a beads-on-a-string-type pattern.

RAD21 may act as the limiting factor for cohesin formation so that up-regulation of RAD21 leads to an increased pool of cohesin. RAD21 may promote cohesin loading on chromatin and thus bias the loading/unloading balance of cohesin for excessive extrusion of chromatin.





□ FM3VCF: A Software Library for Accelerating the Loading of Large VCF Files in Genotype Data Analyses

>> https://www.biorxiv.org/content/10.1101/2023.06.25.546413v1

FM3VCF (fast M3VCF) can convert VCF files into the exclusive data format of MINIMAC4, M3VCF, and efficiently read and parse data from VCF files. In comparison to m3vcftools, FM3VCF is approximately 20 times faster for compressing VCF files to M3VCF format.

The compression task using m3veftools involves three main steps: reading and parsing the VCF file data, compressing and converting the VCF file records to M3 VCF file records, and writing the resulting data into the M3VCF file.

FM3VCF separates the Read, Compress, and Write processes and assigns them to different threads, enabling the three compression steps to be completed in parallel across multiple CPU threads.





□ nf-core/marsseq: systematic pre-processing pipeline for MARS-seq experiments

>> https://www.biorxiv.org/content/10.1101/2023.06.28.546862v1

Mars-seq pipeline is straightforward to execute and involves two main steps. First, the building of the necessary reference indexes for a designated genome. The pipeline aligns the raw reads and generates a count matrix that is then utilized for further downstream analysis.

MARS-seq is a paired-end method where read 1 consists of a left adapter, a pool barcode and cDNA. Read 2 contains a cell barcode and a UMI. To mimic the 10X format, they merge PB, CB and UMI to generate R1 and move the trimmed cDNA to R2.





□ KG-Hub - Building and Exchanging Biological Knowledge Graphs

>> https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btad418/7211646

KG-Hub, a platform that enables standardized construction, exchange, and reuse of knowledge graphs. Features include a simple, modular extract-transform-load (ETL) pattern for producing graphs compliant with Biolink Model, easy integration of any OBO ontology.

All graphs in KG-Hub are represented as directed, heterogeneous property graphs. KG-Hub allows reuse of transformed data across different projects. Each KG project produces a subgraph representing the data from each of the upstream sources that it ingests and transforms.





□ Varda Space Industries

>> https://twitter.com/vardaspace/status/1674871004810858496

Over the last day, for the first time ever, orbital drug processing happened outside of a government-run space station

Our crystallization of Ritonavir appears to have been nominal

This is our first step in commercializing microgravity and building an industrial park in LEO



□ To Find Life in the Universe, Find the Computation

>> https://comdig.unam.mx/2023/06/30/to-find-life-in-the-universe-find-the-computation/





□ StarTal

>> https://twitter.com/startalkradio/status/1674817357678624779

NASA just released Webb’s first image of Saturn 🪐





□ SaseR: Juggling offsets unlocks RNA-seq tools for fast scalable differential usage, aberrant splicing and expression analyses.

>> https://www.biorxiv.org/content/10.1101/2023.06.29.547014v1

An unbiased and fast algorithm for parameter estimation to
assess aberrant expression and splicing that scales better to the large number of latent covariates that are typically needed in studies on rare disease with large cohorts.

saseR (Scalable Aberrant Splicing and Expression Retrieval), vastly outperforms existing SOTA tools as DEXSeg, OUTRIDER, OutSingle and FRASER in terms of computational speed and scalability. More importantly, they dramatically boost the performance for aberrant splicing.





□ An Atlas of Variant Effects to understand the genome at nucleotide resolution

>> https://genomebiology.biomedcentral.com/articles/10.1186/s13059-023-02986-x

MAVEs are a rapidly growing family of methods that involve mutagenesis of a DNA-encoded protein or regulatory element followed by a multiplexed assay for some aspect of function.

Compiling a complete Atlas of Variant Effects for all 20,000 human genes, not to mention potentially hundreds of thousands of noncoding regulatory elements, will require an international collaborative effort involving thousands of researchers, clinicians and technologists.





□ scARE: Attribution Regularization for Single Cell Representation Learning

>> https://www.biorxiv.org/content/10.1101/2023.07.05.547784v1

scARE, a novel end-to-end generative deep learning model, amplifies model sensitivity to a preselected subset of features while minimizing others. scARE incorporates an auxiliary attribution loss term during model training.

scARE uncovers subclusters associated with the expression patterns of two cellular pathway genes, and it optimizes the model training procedure by leveraging time-points metadata.





□ Spontanously breaking of symmetry in overlapping cell instance segmentation using diffusion models

>> https://www.biorxiv.org/content/10.1101/2023.07.07.548066v1

As pixel-level predictors, such as UNet and Cellpose, assign individual pixels to instance masks, these methods cannot be used for overlapping data.

This diffusion model split approach achieves approximately the same score as cellpose, thus demonstrating the same improvement over Mask-R-CNN, but with a model that generalizes to overlapping cells.





□ FRIME: Breaking Down Cell-Free DNA Fragmentation: A Markov Model Approach

>> https://www.biorxiv.org/content/10.1101/2023.07.06.547953v1

FRIME (Fragmentation, Immigration, and Exit), a Markovian model that captures three leading mechanisms governing cfDNA fragmentation. The FRIME model enables the simulation of cfDNA fragment profiles by sampling from the stationary distribution of FRIME processes.

FRIME generates fragment profiles similar to those observed in liquid biopsies and provide insight into the underlying biological mechanisms driving the fragmentation dynamics.





□ miraculix: Accelerated computations for iterative-solver techniques in single-step BLUP models

>> https://www.biorxiv.org/content/10.1101/2023.07.06.547949v1

As an extension to the miraculix package, they have developed tailored solutions for the computation of genotype matrix multiplications, a critical bottleneck when iteratively solving equation systems associated with single-step models.

solved the equation systems associated with the ssSNPBLUP and sGTABLUP models with the program hpblup, a PC-based solver used by the software MiXBLUP 3.1, which links against the miraculix library and toggles the use of the novel implementation through an option.





□ metaMDBG: Efficient High-Quality Metagenome Assembly from Long Accurate Reads using Minimizer-space de Bruijn Graphs

>> https://www.biorxiv.org/content/10.1101/2023.07.07.548136v1

metaMDBG, a method that takes the principle of minimizer space assembly. They also designed a highly efficient multi-k' approach, where the length of k'-min-mers is iteratively increased whilst feeding back the results of the last round of assembly.

The universal minimizers, which are k-mers that map to an integer below a fixed threshold, in each read are first identified. Each read is thus represented as an ordered list of the selected minimizers, denoted a minimizer-space read.