jfresnedo / genome-annotation

Annotation of the Almond genome using MAKER container

Genome Annnotation: Gene prediction & functional annotation

Remarks

• A beginner's guide on on genome assembly and annotation can be found here.
• Download the MAKER container using this script.
• Analysis will be carried out on the Ohio Supercomputer Center. Make sure to name your working directory contained in /fs/scratch/PAS1582/ with the same name as your OSC user name. For instance, if my OSC username is osu123, then have your working directory as /fs/scratch/PAS1582/osu123
• You will be annotating one Scaffold from the Almond genome assembly.

MAKER : A summary of steps

The following is a summary of steps involved in genome annotation using MAKER pipeline. Note that prior identification of repeat elements is essential for genome sequence masking, before predicting genes. This step has already been run, and we will therefore not cover in class. However, interested parties can use this script for repeat identification (uses RepeatMasker).

1.0 Initial MAKER analysis (round 1 training):

Gene model prediction based on evidence (expression data, protein sequence data, repeat annotation). Resulting GFF used to train SNAP gene finder. Code: 1.0_annotation_training.sh

1.1. Train SNAP (round 1)

Initial training of SNAP gene finder. This generates an HMM file/classifier for the first round of gene prediction using SNAP and AUGUSTUS gene finders, next section. Code: 1.1_train_SNAP_round1.sh

2.0. Round 1 of maker prediction

First iteration of gene predictions employing a bootstrap approach. Predict gene models using SNAP and AUGUSTUS gene finders, using the HMM classifier generated in section 1.1 above. The resulting GFFs will be used for re-training the HMM classifier. Code: 2.0_annotation_prediction_1.sh

2.1. Train SNAP (round 2)

Round 2 of SNAP training. Code: 2.1_trains_SNAP_round2.sh Subsequent steps are iterations of predicting models and training HMM. A total of 3 training iterations recommended (to prevent overfitting).

3.0. Round 2 of maker prediction

Code: 3.0_annotation_prediction_2.sh

3.1. Train SNAP (round 3)

Final iteration for training the HMM. Code: 3.1_train_SNAP_round3.sh

4.0. Round 3 of maker prediction

Final round of MAKER gene model predictions. Code: 4.0_annotation_prediction_3.sh

5.0. Functional annotation:

5.1. Uniprot/SwissProt blastp

Blast the MAKER generated protein sequences to UniProt/SwissProt with blastp, and add Uniprot results/functional annotations to the maker GFF3 file as a 'Notes' attribute. Code: 5.1_uniprot.sh

5.2. InterProScan

Search Pfam database for protein domains in the predicted proteins; and update the MAKER generated GFF3 file with the InterProScan results. This is important for identifying predicted gene models lacking evidence (AED=1) but containing Pfam protein domains in their protein sequence. Pfam search also provides GO terms, hence additional functional annotation. Code: 5.2_interProScan.sh

6.0. Final gene models

Filter gene models based on domain content (pfam) and evidence support. Visualize models on a genome browser such as Jbrowse.

Software

A) Maker bioconda installation:

First install conda (if not installed). On Linux, run:

wget	https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
sh	Miniconda3-latest-Linux-x86_64.sh

Follow the instructions in the installer. If you encounter problems, refer to the Miniconda documentation. Add conda to PATH environment:

cd

nano	.bashrc

Add the following to the opened .bashrc:

export	PATH=$PATH:/path/to/miniconda3/bin

Close .bashrc (CTRL +X) Then run:

source	.bashrc
conda	config	--add	channels	defaults
conda	config	--add	channels	bioconda
conda	config	--add	channels	conda-forge

Install MAKER NOTE: It seems MAKER uses Python2.7, not 3, therefore downgrade the Python in miniconda. Run:

conda	install	python=2.

Follow prompts, accept all. Alternativey, switch between Python 2 and environments, see here. Install MAKER:

conda	install	maker

B) Maker (bio)container installation:

Make sure docker is installed on your machine. On your machine, run:

docker	pull	quay.io/biocontainers/maker:<tag>

see maker/tags for valid values for For instance:

docker	pull	quay.io/biocontainers/maker:2.31.10--pl526_13

List docker iamges on your machine:

docker	images

Run the image (create a container)

docker	run	-d	-it	<image	ID	#>	/bin/bash

List containers

docker	ps

Rename container

docker	commit	<container	ID>	wilberzach/maker:biocontaine
rversion

Upload to dockerhub repository

docker	login	-u	wilberzach	-p	<dockerhub	passwd>
docker	push	wilberzach/maker:biocontainerversion

Create/dowload a singularity image in OSC (Linux System). Since the image is large, run an interactive batch for downloading image and creating a Singularity equivalent:

qsub	-l	nodes= 1 :ppn= 10	 -l	walltime= 00 : 45 : 00 	-A	<PROJECT	ID>

When the job starts running, run:

singularity	pull	maker_biocontainer.sif	docker://wilberzach/maker:biocontainerversion