GeMoSeq-Examples - Revision history

Grau at 14:52, 19 November 2025

2025-11-19T14:52:49Z

← Older revision		Revision as of 14:52, 19 November 2025
Line 43:		Line 43:

	The Analyzer tools outputs several metrics, including CDS sensitivity and precision, which are in this case 72.13 and 90.67, respectively.		The Analyzer tools outputs several metrics, including CDS sensitivity and precision, which are in this case 72.13 and 90.67, respectively.




	== Using multiple libraries in combination with homology-based gene prediction ==		== Using multiple libraries in combination with homology-based gene prediction ==
Line 98:		Line 101:

	Here, we assume that for the homology-based predictions, we also obtain a loosely and a strictly filtered prediction file as <code>gemoma_loose/filtered_predictions.gff</code> and <code>gemoma_strict/filtered_predictions.gff</code>, respectively.		Here, we assume that for the homology-based predictions, we also obtain a loosely and a strictly filtered prediction file as <code>gemoma_loose/filtered_predictions.gff</code> and <code>gemoma_strict/filtered_predictions.gff</code>, respectively.


	=== Merging RNA-seq-based and homology-based predictions ===		=== Merging RNA-seq-based and homology-based predictions ===

Grau: /* Merging RNA-seq-based and homology-based predictions */

2025-11-19T14:50:20Z

Merging RNA-seq-based and homology-based predictions

← Older revision		Revision as of 14:50, 19 November 2025
Line 106:		Line 106:
	java -jar GeMoSeq-1.2.3.jar merge g=gemoma_loose/filtered_predictions.gff GeMoSeq=gemoseq_loose/filtered_predictions.gff m=annotate GeMoMa-strict=gemoma_strict/filtered_predictions.gff GeMoSeq-strict=gemoseq_strict/filtered_predictions.gff l=false outdir=final		java -jar GeMoSeq-1.2.3.jar merge g=gemoma_loose/filtered_predictions.gff GeMoSeq=gemoseq_loose/filtered_predictions.gff m=annotate GeMoMa-strict=gemoma_strict/filtered_predictions.gff GeMoSeq-strict=gemoseq_strict/filtered_predictions.gff l=false outdir=final

	where parameter <code>m</code> specifies that merged predictions should be annotated with their confidence (high, medium) and <code>l=false</code> indicates that low-confidence predictions should be excluded from the output. After running the Merge tool, final predictions can be found in <code>final/Merged_Predictions.gff</code>.		where parameter <code>m=annotate</code> specifies that merged predictions should be annotated with their confidence (high, medium) and <code>l=false</code> indicates that low-confidence predictions should be excluded from the output. After running the Merge tool, final predictions can be found in <code>final/Merged_Predictions.gff</code>.

Grau: /* Applying GeMoSeq to individual libraries */

2025-11-19T14:48:24Z

Applying GeMoSeq to individual libraries

← Older revision		Revision as of 14:48, 19 November 2025
Line 56:		Line 56:

	Assuming that the mapped reads are stored in directories named by the accession of the sequencing run, namely <code>mappings/<accession>/Aligned.sortedByCoord.out.bam</code>, e.g., <code>mappings/DRR377420/Aligned.sortedByCoord.out.bam</code>, we need to call GeMoSeq for each of these mapping files independently. This can be achieved by a (bash) for loop over all BAM files or in parallel on a compute cluster using a scheduling system like slurm.		Assuming that the mapped reads are stored in directories named by the accession of the sequencing run, namely <code>mappings/<accession>/Aligned.sortedByCoord.out.bam</code>, e.g., <code>mappings/DRR377420/Aligned.sortedByCoord.out.bam</code>, we need to call GeMoSeq for each of these mapping files independently. This can be achieved by a (bash) for loop over all BAM files or in parallel on a compute cluster using a scheduling system like slurm.

			An individual GeMoSeq run would look like

			java -jar GeMoSeq-1.2.3.jar gemoseq g=NbLab360.genome.fasta m=mappings/<accession>/Aligned.sortedByCoord.out.bam s=<strandedness> threads=6 outdir=gemoseq/<accession>

			where <code><accession></code> refers to the accession and <code><strandedness></code> refers to the strand orientation of the current library.

	For instance, library DRR377420 has strand orientation <code>FR_FIRST_STRAND</code> (see above), and the corresponding GeMoSeq call would be		For instance, library DRR377420 has strand orientation <code>FR_FIRST_STRAND</code> (see above), and the corresponding GeMoSeq call would be

Grau: /* Using multiple libraries in combination with homology-based gene prediction */

2025-11-19T14:45:41Z

Using multiple libraries in combination with homology-based gene prediction

← Older revision		Revision as of 14:45, 19 November 2025
Line 46:		Line 46:
	== Using multiple libraries in combination with homology-based gene prediction ==		== Using multiple libraries in combination with homology-based gene prediction ==

	We use a [~~https~~://~~bioweb01~~.~~qut~~.~~edu.au~~/~~N.bethamiana~~/NbLab360.genome.fasta.gz recently published genome] of a ''Nicotiana benthamiana'' lab strain ([https://doi.org/10.1038/s41477-023-01489-8]) to illustrate how GeMoSeq can be used to re-annotate a genome based on multiple sequencing libraries and in combination with the homology-based gene prediction of [[GeMoMa]].		We use a [http://www.waterhouse-lab.com/Benth_download_2025/NbLab360.genome.fasta.gz recently published genome] of a ''Nicotiana benthamiana'' lab strain ([https://doi.org/10.1038/s41477-023-01489-8]) to illustrate how GeMoSeq can be used to re-annotate a genome based on multiple sequencing libraries and in combination with the homology-based gene prediction of [[GeMoMa]].

	In the [https://doi.org/10.1101/2025.02.27.640589 GeMoSeq preprint], we sampled 319 sequencing libraries for ''N. benthamiana'' for this purpose, and we provide the corresponding [https://www.ebi.ac.uk/ena/browser/home ENA] accession table [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx for download].		In the [https://doi.org/10.1101/2025.02.27.640589 GeMoSeq preprint], we sampled 319 sequencing libraries for ''N. benthamiana'' for this purpose, and we provide the corresponding [https://www.ebi.ac.uk/ena/browser/home ENA] accession table [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx for download].

	All of these libraries need to be downloaded, mapped to the genome sequence, and strandedness needs to be determined. For mapping, we follow the same procedure as described for the GeMoSeq example run above. To determine strandedness, we use [https://subread.sourceforge.net/featureCounts.html featureCounts] applied to the previous [genome annotation] with flags for either strand orientation and choose the one with the larger number of assigned reads, or unstranded if both yield roughly equal numbers of assigned reads.		All of these libraries need to be downloaded, mapped to the genome sequence, and strandedness needs to be determined. For mapping, we follow the same procedure as described for the GeMoSeq example run above. To determine strandedness, we use [https://subread.sourceforge.net/featureCounts.html featureCounts] applied to the previous [http://www.waterhouse-lab.com/Benth_download_2025/NbLab360.v103.gff3.gz genome annotation] with flags for either strand orientation and choose the one with the larger number of assigned reads, or unstranded if both yield roughly equal numbers of assigned reads.
	The strand orientation that has been determined by this procedure is listed in column "type" of the [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx accession table].		The strand orientation that has been determined by this procedure is listed in column "type" of the [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx accession table].

Grau: /* Running GeMoSeq */

2025-11-19T14:44:06Z

Running GeMoSeq

← Older revision		Revision as of 14:44, 19 November 2025
Line 34:		Line 34:

	Here, we use the parameter <code>s=FR_FIRST_STRAND</code> to specify that this is a stranded library and set the number of threads to 6. The output directory is set to gemoseq.		Here, we use the parameter <code>s=FR_FIRST_STRAND</code> to specify that this is a stranded library and set the number of threads to 6. The output directory is set to gemoseq.
	At execution, GeMoSeq creates a temporary file with predictions in the working directory, which is later copied to the specified output directory together with a protocol of the GeMoSeq run. In the output directory you find the final predictions as a file <code>Transcript_Predictions.gff3</code>.		At execution, GeMoSeq creates a temporary file with predictions in the working directory, which is later copied to the specified output directory together with a protocol of the GeMoSeq run. In the output directory, you will find the final predictions as a file <code>Transcript_Predictions.gff3</code>.

	=== Evaluating prediction performance ===		=== Evaluating prediction performance ===

Grau: /* GeMoSeq example run */

2025-11-19T14:42:35Z

GeMoSeq example run

← Older revision		Revision as of 14:42, 19 November 2025
Line 3:		Line 3:
	We use a data set for yeast for an example run of GeMoSeq. As input of GeMoSeq, we need a reference genome and a library of reads mapped to that reference.		We use a data set for yeast for an example run of GeMoSeq. As input of GeMoSeq, we need a reference genome and a library of reads mapped to that reference.
	Here, we use the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.fna.gz ''S. cerevisiae'' S288C reference genome R64] available from NCBI. For later evaluation of GeMoSeq prediction results, we also download the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.gff.gz reference annotation].		Here, we use the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.fna.gz ''S. cerevisiae'' S288C reference genome R64] available from NCBI. For later evaluation of GeMoSeq prediction results, we also download the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.gff.gz reference annotation].
	As a sequencing library we use the library with accession SRR8552464 available from the [https://www.ebi.ac.uk/ena/browser/home European Nucleotide archive] or [https://www.ncbi.nlm.nih.gov/sra NCBI Sequence Read Archive].		As a sequencing library, we use the library with accession SRR8552464 available from the [https://www.ebi.ac.uk/ena/browser/home European Nucleotide archive] or [https://www.ncbi.nlm.nih.gov/sra NCBI Sequence Read Archive].

	=== Mapping ===		=== Mapping ===
Line 43:		Line 43:

	The Analyzer tools outputs several metrics, including CDS sensitivity and precision, which are in this case 72.13 and 90.67, respectively.		The Analyzer tools outputs several metrics, including CDS sensitivity and precision, which are in this case 72.13 and 90.67, respectively.




	== Using multiple libraries in combination with homology-based gene prediction ==		== Using multiple libraries in combination with homology-based gene prediction ==

Grau at 16:09, 18 November 2025

2025-11-18T16:09:25Z

← Older revision		Revision as of 16:09, 18 November 2025
Line 1:		Line 1:


	== GeMoSeq example run ==		== GeMoSeq example run ==

Grau at 16:09, 18 November 2025

2025-11-18T16:09:13Z

← Older revision		Revision as of 16:09, 18 November 2025
Line 1:		Line 1:


	== GeMoSeq example run ==		== GeMoSeq example run ==

Grau: /* Joining individual predictions */

2025-11-18T16:08:39Z

Joining individual predictions

← Older revision		Revision as of 16:08, 18 November 2025
Line 71:		Line 71:
	=== Joining individual predictions ===		=== Joining individual predictions ===

	To yield a joint GeMoSeq prediction across all libraries, we join the individual predictions using the [[GeMoMa-Docs#GeMoMa_Annotation_Filter GeMoMa Annotation Filter]] (GAF) tool of [[GeMoMa]], which also comes packaged with the GeMoSeq JAR file.		To yield a joint GeMoSeq prediction across all libraries, we join the individual predictions using the [[GeMoMa-Docs#GeMoMa_Annotation_Filter\| GeMoMa Annotation Filter]] (GAF) tool of [[GeMoMa]], which also comes packaged with the GeMoSeq JAR file.
	For future merging with homology-based GeMoMa predictions, we run the GAF tool twice, once with looser and once with stricter filtering criteria.		For future merging with homology-based GeMoMa predictions, we run the GAF tool twice, once with looser and once with stricter filtering criteria.

Line 87:		Line 87:

	and obtain joint predictions in <code>gemoseq_strict/filtered_predictions.gff</code>.		and obtain joint predictions in <code>gemoseq_strict/filtered_predictions.gff</code>.


	=== Homology-based gene prediction ===		=== Homology-based gene prediction ===

Grau: Created page with " == GeMoSeq example run == We use a data set for yeast for an example run of GeMoSeq. As input of GeMoSeq, we need a reference genome and a library of reads mapped to that reference. Here, we use the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.fna.gz ''S. cerevisiae'' S288C reference genome R64] available from NCBI. For later evaluation of GeMoSeq prediction results, we also download the [https://ftp.ncbi.nlm..."

2025-11-18T16:05:01Z

Created page with " == GeMoSeq example run == We use a data set for yeast for an example run of GeMoSeq. As input of GeMoSeq, we need a reference genome and a library of reads mapped to that reference. Here, we use the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.fna.gz ''S. cerevisiae'' S288C reference genome R64] available from NCBI. For later evaluation of GeMoSeq prediction results, we also download the [https://ftp.ncbi.nlm..."

New page

== GeMoSeq example run ==

We use a data set for yeast for an example run of GeMoSeq. As input of GeMoSeq, we need a reference genome and a library of reads mapped to that reference.
Here, we use the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.fna.gz ''S. cerevisiae'' S288C reference genome R64] available from NCBI. For later evaluation of GeMoSeq prediction results, we also download the [https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/146/045/GCF_000146045.2_R64/GCF_000146045.2_R64_genomic.gff.gz reference annotation].
As a sequencing library we use the library with accession SRR8552464 available from the [https://www.ebi.ac.uk/ena/browser/home European Nucleotide archive] or [https://www.ncbi.nlm.nih.gov/sra NCBI Sequence Read Archive].

=== Mapping ===

''If you just want to test GeMoSeq, you can skip this step and proceed with the download provided at the end of this section.''

Before proceeding with GeMoSeq, these reads first need to be mapped to the reference genome. For yeast and other eukaryotes, we should use a splicing-aware mapping tool. Our current favourite is [https://github.com/alexdobin/STAR STAR] also available from [http://bioconda.github.io/recipes/star/README.html bioconda].

First, we need to index the reference genome with

STAR --runMode genomeGenerate --genomeDir yeast_index --genomeFastaFiles GCF_000146045.2_R64_genomic.fna --genomeSAindexNbases 10

Here, we specify <code>--genomeSAindexNbases 10</code> since the yeast genome is rather small.

Assuming reads have been stored in files <code>SRR8552464_1.fastq.gz</code> and <code>SRR8552464_2.fastq.gz</code>, we run STAR with the following command

STAR --genomeDir yeast_index --readFilesIn SRR8552464_1.fastq.gz SRR8552464_2.fastq.gz --readFilesCommand gunzip -c --outFileNamePrefix star/SRR8552464 --alignIntronMax 20000 --alignMatesGapMax 20000 --outSAMtype BAM SortedByCoordinate

Here, we limit the maximum intron size to 20000, since larger introns do not exist in yeast, but STAR tends to produce spurios long-split mappings, and let STAR output an already coordinate-sorted BAM file.

For convenience, we provide the resulting mapping file [http://www.jstacs.de/downloads/GeMoSeq/SRR8552464.bam for download].

=== Running GeMoSeq ===

The resulting mapping file and the reference genome used for mapping are input of GeMoSeq in the next step. For the next steps, we assume that the BAM file has been moved to a file <code>SRR8552464.bam</code> in the working directory (or has been downloaded from the provided link).

We leave most arguments at their defaults, calling GeMoSeq with

java -jar GeMoSeq-1.2.3.jar gemoseq g=GCF_000146045.2_R64_genomic.fna m=SRR8552464.bam s=FR_FIRST_STRAND threads=6 outdir=gemoseq

Here, we use the parameter <code>s=FR_FIRST_STRAND</code> to specify that this is a stranded library and set the number of threads to 6. The output directory is set to gemoseq.
At execution, GeMoSeq creates a temporary file with predictions in the working directory, which is later copied to the specified output directory together with a protocol of the GeMoSeq run. In the output directory you find the final predictions as a file <code>Transcript_Predictions.gff3</code>.

=== Evaluating prediction performance ===

To evaluate the accuracy of GeMoSeq predictions, we compare the GeMoSeq GFF against the yeast reference annotation. For this purpose, the GeMoSeq Jar file provides the [[GeMoMa]] Analyzer tool, which can be run with

java -jar GeMoSeq-1.2.3.jar Analyzer t=GCF_000146045.2_R64_genomic.gff p=gemoseq/Transcript_Predictions.gff3

The Analyzer tools outputs several metrics, including CDS sensitivity and precision, which are in this case 72.13 and 90.67, respectively.

== Using multiple libraries in combination with homology-based gene prediction ==

We use a [https://bioweb01.qut.edu.au/N.bethamiana/NbLab360.genome.fasta.gz recently published genome] of a ''Nicotiana benthamiana'' lab strain ([https://doi.org/10.1038/s41477-023-01489-8]) to illustrate how GeMoSeq can be used to re-annotate a genome based on multiple sequencing libraries and in combination with the homology-based gene prediction of [[GeMoMa]].

In the [https://doi.org/10.1101/2025.02.27.640589 GeMoSeq preprint], we sampled 319 sequencing libraries for ''N. benthamiana'' for this purpose, and we provide the corresponding [https://www.ebi.ac.uk/ena/browser/home ENA] accession table [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx for download].

All of these libraries need to be downloaded, mapped to the genome sequence, and strandedness needs to be determined. For mapping, we follow the same procedure as described for the GeMoSeq example run above. To determine strandedness, we use [https://subread.sourceforge.net/featureCounts.html featureCounts] applied to the previous [genome annotation] with flags for either strand orientation and choose the one with the larger number of assigned reads, or unstranded if both yield roughly equal numbers of assigned reads.
The strand orientation that has been determined by this procedure is listed in column "type" of the [http://www.jstacs.de/downloads/GeMoSeq/N_benthamiana_datasets.xlsx accession table].

=== Applying GeMoSeq to individual libraries ===

Assuming that the mapped reads are stored in directories named by the accession of the sequencing run, namely <code>mappings/<accession>/Aligned.sortedByCoord.out.bam</code>, e.g., <code>mappings/DRR377420/Aligned.sortedByCoord.out.bam</code>, we need to call GeMoSeq for each of these mapping files independently. This can be achieved by a (bash) for loop over all BAM files or in parallel on a compute cluster using a scheduling system like slurm.

For instance, library DRR377420 has strand orientation <code>FR_FIRST_STRAND</code> (see above), and the corresponding GeMoSeq call would be

java -jar GeMoSeq-1.2.3.jar gemoseq g=NbLab360.genome.fasta m=mappings/DRR377420/Aligned.sortedByCoord.out.bam s=FR_FIRST_STRAND threads=6 outdir=gemoseq/DRR377420

where we, again, choose the name of the output directory based on the accession of the sequencing run. Within that directory, each GeMoSeq run will produce a file <code>Transcript_Predictions.gff3</code> with the predicted gene models.

After GeMoSeq has been run for all mapping files, the resulting prediction files are located in <code>gemoseq/<accession>/Transcript_Predictions.gff3</code>.

=== Joining individual predictions ===

To yield a joint GeMoSeq prediction across all libraries, we join the individual predictions using the [[GeMoMa-Docs#GeMoMa_Annotation_Filter GeMoMa Annotation Filter]] (GAF) tool of [[GeMoMa]], which also comes packaged with the GeMoSeq JAR file.
For future merging with homology-based GeMoMa predictions, we run the GAF tool twice, once with looser and once with stricter filtering criteria.

For the looser filtering, we specify

java -jar GeMoSeq-1.2.3.jar GAF p=DRR377420 g=gemoseq/DRR377420/Transcript_Predictions.gff3 ... p=SRR29507616 g=gemoseq/SRR29507616/Transcript_Predictions.gff3 f="start=='M' and stop=='*' and sumWeight>7 and (isNaN(tie) or tie==1)" atf="sumWeight>23" mnotpg=20 tf=true outdir=gemoseq_loose

where <code>...</code> indicates that the arguments <code>p</code> and <code>g</code> need to be repeated for each of the individual predictions.
The filtering criteria (parameter <code>f</code>) specify that each tranferred transcript must have a proper start and stop codon, must appear in more than 7 (of 319) individual predictions and may either be a single-exon transcript (<code>isNaN(tie)</code>) or have perfect support for all splice sites (<code>tie==1</code>). The parameter <code>atf</code> defines that alternative transcripts are only transferred, if they occur in at least 24 of 319 libraries, parameter <code>mnotpg</code> specifies the maximum number of transcripts per gene, and <code>tf=true</code> indicates that additional features (i.e., exons + CDS) should be tranferred to the final prediction.
After the GAF run, the joint predictions are stored in <code>gemoseq_loose/filtered_predictions.gff</code>.

For the stricter filtering, we use adapted parameters

java -jar GeMoSeq-1.2.3.jar GAF p=DRR377420 g=gemoseq/DRR377420/Transcript_Predictions.gff3 ... p=SRR29507616 g=gemoseq/SRR29507616/Transcript_Predictions.gff3 f="start=='M' and stop=='*' and sumWeight>30 and (isNaN(tie) or tie==1)" atf="sumWeight>70" mnotpg=20 tf=true outdir=gemoseq_strict

and obtain joint predictions in <code>gemoseq_strict/filtered_predictions.gff</code>.

=== Homology-based gene prediction ===

For homology-based gene prediction, we use [[GeMoMa]] for the reference species ''A. thaliana'', ''N. tabacum'', ''S. lycopersicum'', and two previous ''N. benthamiana'' genome versions (1.0.1 and 2.6.1).
These predictions are also joined using the GAF tool.

For details about using GeMoMa for homology-based gene prediction, please refer to [[GeMoMa-Docs]].

Here, we assume that for the homology-based predictions, we also obtain a loosely and a strictly filtered prediction file as <code>gemoma_loose/filtered_predictions.gff</code> and <code>gemoma_strict/filtered_predictions.gff</code>, respectively.

=== Merging RNA-seq-based and homology-based predictions ===

We finally merge the four previously obtained prediction files (loose and strict for GeMoSeq and GeMoMa, respectively) using the GeMoSeq Merge tool. We call

java -jar GeMoSeq-1.2.3.jar merge g=gemoma_loose/filtered_predictions.gff GeMoSeq=gemoseq_loose/filtered_predictions.gff m=annotate GeMoMa-strict=gemoma_strict/filtered_predictions.gff GeMoSeq-strict=gemoseq_strict/filtered_predictions.gff l=false outdir=final

where parameter <code>m</code> specifies that merged predictions should be annotated with their confidence (high, medium) and <code>l=false</code> indicates that low-confidence predictions should be excluded from the output. After running the Merge tool, final predictions can be found in <code>final/Merged_Predictions.gff</code>.