Mingfu Shao

This paper improves the minimum flow decomposition algorithm for bioinformatics tasks by jointly modeling equations in flow networks, achieving faster and more accurate results.

TENNIS predicts unannotated isoforms and refines existing annotations using an evolution-based model of alternative splicing, validated by deep long-read RNA-seq.

EquiRep accurately detects tandem repeats from error-prone sequences by refining unit estimation and reconstructing repeat units through a weighted graph approach.

Amaranth improves single-cell transcript assembly by discriminatively modeling UMI and internal reads, outperforming existing assemblers on Smart-seq3 datasets.

MELO-ED improves edit distance computation by using multi-embeddings to approximate similarity between biological sequences efficiently.

The paper introduces CircPlex, a method to accurately reconstruct circRNAs from complex rolling circular long reads by addressing repeat pattern complexities, improving detection sensitivity and uncovering novel isoforms.

Telos improves transcript assembly by accurately detecting TSSs and TESs and ranking transcripts, outperforming existing methods across multiple RNA-seq datasets.

This paper introduces LLMED, a genomic large language model-based embedding method that improves edit distance approximation and similar sequence search accuracy.

This paper introduces SubseqHash2, an improved subsequence-based seeding method that achieves faster computation while maintaining high accuracy for error-prone long reads.

Beaver improves transcriptome assembly at single-cell resolution by combining a fragment graph with random forest models, outperforming existing assemblers in precision.

This paper introduces SubseqSketch, an alignment-free method for sequence similarity estimation that uses random subsequences to efficiently approximate edit distances, showing effectiveness in benchmark tasks.

This paper introduces a SAT formulation for computing the DCJ distance between genomes with duplicates, showing it outperforms ILP in speed and scalability.

TERRACE improves circRNA assembly accuracy by using a splice graph and dynamic programming, outperforming existing methods in sensitivity and precision.

The paper investigates how transcript annotations affect assembly results and develops a tool to reduce intron retention bias in transcript assembly.

Scallop2 improves transcript assembly accuracy for multiple-end RNA-seq data by efficiently bridging gaps and reducing false positives.

This paper introduces algorithms to reconstruct full RNA-seq fragments from paired-end reads, improving transcriptome assembly and expression quantification.

The paper introduces a new seeding framework for read mapping that reduces seed frequencies and improves efficiency by using context-aware seeds with confidence radii.

Coral improves transcript assembly by accurately bridging paired-end RNA-seq reads, achieving 7.5% and 11.2% adjusted precision improvements on GTEx data.

This paper evaluates how transcript assembly improves transcriptome analysis using long-read sequencing, showing Scallop-LR identifies more known transcripts and novel isoforms than Iso-Seq Analysis.

SQUID detects transcriptomic structural variations from RNA-seq by improving accuracy on simulation data and identifying novel non-fusion-gene TSVs in TCGA samples.

Scallop improves transcript assembly by preserving phasing paths and reducing coverage deviation, outperforming existing assemblers in multi-exon and lowly expressed transcript reconstruction.

DeepBound identifies transcript boundaries from RNA-seq reads using deep convolutional neural fields, outperforming existing methods on simulated and biological datasets.

The paper introduces an efficient heuristic for decomposing a flow into the minimum number of paths, showing significant improvement over previous methods in both simulated and transcript assembly scenarios.