Hirak Sarkar

SPaRTA enables non-experts to query spatial-omics data via natural language, achieving 63-77% accuracy with fine-tuning on cancer tissue datasets.

This study shows that caloric restriction before tumor induction delays cancer cachexia by reducing systemic inflammation and preserving lean mass.

FEAST generates synthetic spatial transcriptomics data with tunable spatial and transcriptional variation for benchmarking and 3D reconstruction.

This paper introduces SIID, an algorithm that integrates spatial transcriptomics data from different platforms to impute missing gene expression and deconvolve mixed cell-type signals.

The paper identifies iron-recycling macrophages in bone metastases that support tumor growth and anemia by hijacking erythropoiesis and enabling tumor cell erythroblast mimicry.

This paper introduces GASTON, an interpretable deep learning method to map spatial gene expression topography, identifying gradients and domains in spatial transcriptomics data.

DifferentialRegulation identifies differentially regulated genes by analyzing unspliced mRNA abundance across experimental conditions using a Bayesian hierarchical model.

This paper introduces Numbat, a method that uses haplotype information to improve detection of copy number variations from scRNA-seq data, revealing subclonal complexity in cancer.

Airpart identifies cell-type-specific allelic imbalance from single-cell RNA-seq data using statistical models and provides interpretable partitions of genes and cells.

Alevin-fry provides a fast, memory-efficient, and accurate method for quantifying single-cell and single-nucleus RNA-seq data, including RNA velocity inputs.

This paper revises a benchmark to compare kallisto-bustools and alevin-fry, showing alevin-fry is faster and more memory-efficient when configured similarly to kallisto-bustools.

The paper confirms that considering unexpected fragment origins improves transcript quantification accuracy in RNA-seq simulations and shows that Srivastava et al.'s methodology enhances this accuracy.

The paper introduces a method to compress quantification uncertainty in scRNA-seq data by storing mean and variance instead of full inferential replicates, improving computational efficiency without loss of performance.

The paper evaluates how alignment and mapping methods affect transcript quantification accuracy and introduces a new method called selective alignment.

The paper introduces a Bayesian framework for improving dscRNA-seq quantification by sharing information across cells and modalities, enhancing gene abundance estimates for multi-mapping reads.

Terminus groups transcripts based on inferential uncertainty to enable robust transcript-level or group-level analysis from RNA-seq data.

The paper introduces a space and time-efficient index for the compacted colored de Bruijn graph, balancing speed and space usage for k-mer lookup and pattern matching.

The paper introduces a selective-alignment algorithm to improve transcript-level quantification accuracy while maintaining efficiency, resolving complex mapping scenarios and reducing spurious alignments.

Quark compresses RNA-seq data using a semi-reference-based approach, achieving better compression than reference-free methods while allowing reference-free decompression.

RapMap introduces a quasi-mapping approach to efficiently align RNA-seq reads to transcriptomes, improving speed and accuracy over existing tools.