LoRTIA Plus: a chemistry-agnostic, feature-first software package for long-read transcriptome annotationItem

Read full text Summarise with VIRAT
Authors: Gábor Torma
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Update – September 2026The Figshare dataset has been updated with four new revision-stage benchmark archives supporting the expanded validation and reproducibility analyses of the revised LoRTIA Plus manuscript.The newly added archives are:KSHV_revision_with_StringTie2_LAPA.rar – contains the revised KSHV benchmarking workflow, including the expanded comparator set with StringTie2 and LAPA, the consolidated KSHV reference framework, transcript-boundary and transcript-level benchmark inputs, annotator predictions, evaluation outputs, summary tables, and associated reproducibility materials.LRGASP_SQANTI_with_stringtie2_LAPA_revision.rar – contains the revised empirical human LRGASP benchmarking analyses, including TSS/TES boundary benchmarking, SQANTI3-based transcriptome statistics, FSM/ISM transcript-recovery analyses, NIC/NNC and transcript-structure analyses, and the corresponding summary tables, source data, and plotting/reproducibility scripts. The revision-stage comparator set includes StringTie2 and LAPA where applicable.SQANTI_SIM_TKSM_synthetic_benchmark_revision.rar – contains the newly added annotation-independent synthetic benchmark generated from a predefined human transcript truth set using SQANTI-SIM and TKSM. The archive includes the chr17+chr19 reference subset, final SQANTI-SIM truth set, TKSM molecule representation, matched ONT and PacBio synthetic datasets, final alignments, reconstructed transcript models from all six evaluated annotators, evaluation-specific preprocessing files, final SQANTI-SIM outputs, publication-level summary tables, figure source data, and complete reproducibility scripts and documentation.LoRTIA_Plus_JAFFAL_Fusion_Benchmark_v1.0.rar – contains the revision-stage fusion-detection benchmark comparing LoRTIA Plus with JAFFAL using simulated Oxford Nanopore long-read fusion data. The archive includes native and standardized JAFFAL outputs, the final LoRTIA Plus fusion-candidate table, preprocessing and evaluation scripts, benchmark result tables, execution logs, breakpoint-level evaluation outputs, and detailed reproducibility documentation. The original third-party simulation data and reference files are not redistributed; their provenance and regeneration steps are documented in the accompanying README.These revision-stage archives keep the viral, empirical human, independent synthetic, and fusion-detection benchmarking workflows clearly separated while providing the files and documentation required to inspect and reproduce the reported analyses.Dataset descriptionThis dataset contains the benchmarking data, intermediate results, final evaluation outputs, and reproducibility materials generated for the revised LoRTIA Plus manuscript.The dataset supports four complementary benchmarking frameworks used to evaluate long-read transcript annotation and fusion-detection performance across viral, empirical human, independent synthetic, and simulated fusion datasets.1. KSHV benchmarkThe viral benchmark evaluates LoRTIA Plus and comparator transcript annotators using Kaposi’s sarcoma-associated herpesvirus (KSHV) long-read RNA-sequencing data.The revised KSHV benchmark is provided in:KSHV_revision_with_StringTie2_LAPA.rarThe benchmark includes transcript-boundary and transcript-level comparisons against a consolidated KSHV reference set. Transcription start sites (TSSs) and transcription end sites (TESs) were restricted using orthogonal experimental support, including CAGE/RAMPAGE-derived evidence for TSSs and direct-RNA poly(A)-associated evidence for TESs. Transcript structures were evaluated using the corresponding curated reference framework.The revised archive includes the expanded comparator set, including StringTie2 and LAPA where applicable, together with the reference inputs, annotator predictions, benchmark tables, evaluation outputs, and associated reproducibility materials.2. Human LRGASP benchmarkThe empirical human benchmark is provided in:LRGASP_SQANTI_with_stringtie2_LAPA_revision.rarIt uses long-read RNA-sequencing datasets from the LRGASP project and includes multiple cell lines and sequencing chemistries.The revision-stage archive contains the updated workflows and outputs for:TSS and TES boundary benchmarking;SQANTI3-based basic transcriptome statistics;FSM/ISM transcript-recovery analyses;NIC/NNC and transcript-structure analyses;chemistry- and cell-line-resolved comparisons across annotators.The evaluated datasets include H1, H1-derived endoderm, and WTC11 cells generated using ONT and PacBio long-read sequencing protocols, including CapTrap datasets where applicable.The revised comparisons include LoRTIA Plus together with the relevant comparator annotators, including StringTie2, bambu, FLAIR, IsoQuant, NAGATA, and LAPA for boundary-focused analyses.For the human boundary benchmark, active reference subsets were derived from independent reference resources and restricted to loci represented in the evaluated data. These analyses were complemented by transcript-structure benchmarking using SQANTI3.3. Independent SQANTI-SIM/TKSM synthetic benchmarkThe independent synthetic benchmark is provided in:SQANTI_SIM_TKSM_synthetic_benchmark_revision.rarThis benchmark was generated to complement the empirical LRGASP analyses and to address potential annotator dependence associated with active-reference benchmarking.The benchmark was generated using SQANTI-SIM v0.3.0, GENCODE Release 48, and the GRCh38 reference genome, restricted to chromosomes 17 and 19.SQANTI-SIM classif initially classified 37,269 transcript models. A subsequent design equal step generated a common benchmark truth set containing:30,000 transcript models;6,321 genes;90,000 requested reads per platform;exactly three reads per selected truth transcript;a fixed random seed of 42.The final benchmark truth contained 24,212 known/FSM transcript models and 5,788 novel transcript models. The novel subset included ISM, NIC, NNC, genic/genomic, antisense, fusion, and intergenic transcript structures.An explicit three-copy molecule representation was generated from the SQANTI-SIM truth set and converted to TKSM MDF format. Matched ONT and PacBio synthetic datasets were derived from the same predefined truth set.The verified ONT simulation workflow included TKSM poly(A) generation, strand randomization, adapter tagging, and perfect sequence generation, producing 90,000 error-free synthetic reads. The corresponding PacBio synthetic FASTQ and alignment files are included as the authoritative inputs for the PacBio branch.Six transcript annotators were evaluated:LoRTIA Plus;bambu;FLAIR;IsoQuant;NAGATA;StringTie2.The predefined truth set was not available to the reconstruction tools during transcript annotation. Evaluation was performed only after reconstruction using SQANTI-SIM.Evaluation endpoints include gene-level recovery, known/FSM transcripts, novel transcripts, ISM, NIC, NNC, genic/genomic transcripts, antisense transcripts, fusion transcripts, and intergenic transcripts.The synthetic-benchmark archive contains:the chr17+chr19 GENCODE v48 / GRCh38 reference subset;the final SQANTI-SIM truth set;the exact three-copy TKSM MDF representation;simulated ONT and PacBio reads;final ONT and PacBio alignments;reconstructed transcript models from all six annotators;evaluation-specific preprocessing inputs and outputs;final SQANTI-SIM metric tables;publication-level summary tables;figure source data;the final plotting script;scripts and detailed step-by-step reproducibility documentation.Evaluation-specific preprocessing is explicitly documented. StringTie2 records with undefined strand were represented on both strands for SQANTI-SIM compatibility without selecting a favorable strand. PacBio NAGATA identifiers were made chromosome-specific to resolve locally reused identifiers; genomic coordinates and transcript structures were unchanged.All reported LoRTIA Plus synthetic-benchmark results were generated after correcting endpoint-support aggregation so that TSS/TES support was aggregated within the predefined positional wobble before feature qualification. No truth/reference coordinates were used for this correction.4. LoRTIA Plus–JAFFAL fusion-detection benchmarkThe fusion-detection benchmark is provided in:LoRTIA_Plus_JAFFAL_Fusion_Benchmark_v1.0.rarThis revision-stage archive contains the benchmark used to evaluate the fusion-detection functionality of LoRTIA Plus against JAFFAL using simulated Oxford Nanopore long-read fusion data associated with the JAFFAL study.The archive contains:the native JAFFAL jaffa_results.csv output;the JAFFAL execution log;standardized JAFFAL all-call and HighConfidence candidate tables;the final LoRTIA Plus fusion-candidate table used for benchmarking;the script used to convert native JAFFAL output into the common benchmark candidate format;the common LoRTIA Plus–JAFFAL evaluation script;matched-event tables;false-positive and false-negative tables;support-stratified sensitivity outputs;benchmark summary files;breakpoint-error statistics;detailed reproducibility documentation, including the preprocessing and analysis commands used for the benchmark.The benchmark was based on the ONT_fus_sim_85err.fastq.gz simulated long-read fusion dataset and the corresponding truth information from the JAFFAL simulation resources.The original simulated FASTQ, external truth files, GRCh38 reference FASTA, and derived SAM alignment are intentionally not redistributed in this archive. Their provenance, filenames, preprocessing steps, and regeneration commands are documented in the accompanying README.ReproducibilityThe dataset is organized so that the major benchmarking stages can be independe

Fields of Study & Subjects

Genomics and Phylogenetic Studies Single-cell and spatial transcriptomics RNA Research and Splicing

Publication details

Journal / source
Figshare
Publication date
31 March 2026
Language
EN
Document type
Dataset