Behavioral and transcriptomic analysis of age-based polyethism in the pharaoh's ant (Monomorium pharaonis)

Reference based transcriptome assembly and gene expression counts

After filtering adaptor (rename.sh) and low quality sequence (trim.sh), concatenate results (concat.sh) and assemble genome reference using ABYSS (abyss.sh), using a range of k-mers from 51 to 91. k=69 produced the longest N50, and was chosen for future work.

n		n:500	n:N50	min	N80		N50		N20		max		sum
216388	42794	4628	500	5605	16262	36932	246554	283e6	contigs
207026	36001	3980	500	6743	18960	43332	246554	283.7e6	scaffolds

Map transcriptional data to the reference genome using tophat, and infer loci and isoforms using cufflinks (tophat.sh and cuff.sh). After merging all the libraries using cuffmerge, extract reference transcriptome using cufflink's gffread utility.

Now use rsem.sh to map reads to extracted loci, providing a genes to isoforms table based on cufflinks output. This produces files with counts, which are then collected (collect_counts.py and collect_fpkm.py) entered into a SQL database.

Mapping stats for rsem

id  	mapped		percent	library
MP11	4802838 	33.73%	age6
MP3 	4670833 	34.87%	age3
MP5 	5785570 	38.60%	age12
MP9 	7269589 	52.91%	nurse
MP17	7767423 	59.50%	nurse
MP22	11480354	67.10%	groom
MP21	9769906 	69.22%	troph
MP23	10854923	73.04%	age18
MP1 	10664516	77.03%	age0
MP16	11809890	82.97%	groom
MP20	14648253	83.64%	age9
MP19	14522217	84.37%	prot
MP7 	11782737	85.74%	troph
MP14	10697715	86.00%	age18
MP8 	12053556	86.29%	age0
MP2 	11246559	86.58%	age3
MP10	10379494	86.64%	age6
MP12	13243001	86.72%	age15
MP4 	12984328	86.74%	carb
MP13	11114399	86.91%	age15
MP24	16818668	86.98%	prot
MP18	11041174	87.14%	carb
MP6 	12885559	87.29%	age12
MP15	12227812	87.50%	age9

Average: 74.9% +/- 17.5%

Transcriptome annotation and evolutionary rates analysis

Blastx assembled transcriptome vs NR database with evalue 0.00001. Run blast2go on the blast results to obtain GO term annotations. These are collected (collect_go.py) and uploaded to a SQL database

First, we need to get just one locus for every isoform. For this, we just choose the longest transcript.

python choose_isoform.py > ref/longest_isoforms.fa

This produces 22385 loci.

Evolutionary rates vs fire ants (Solenopsis invicta)

Using longest_isoforms.fa conduct reciprocal best blast vs the S. invicta OGSv2.2.3 using blastx and and tblastn with evalue evalue 1e-10. These blast results (which were split into groups of 100 sequences for speed), can also be used to predict protein codeing genes using OrfPredictor:

#predict proteins using blast results
for i in *.fa ; do perl /apps/MikheyevU/sasha/ORFPredictor/OrfPredictor_web3.pl $i `basename $i .fa`.xml 1 both [email protected] 1e-10 `basename $i .fa`_prot.fa test; done

Parse blast data using combine_homologs.py to extract reciprocal best blast hits, and use combine_sequences.py to create fasta files containing the S. invicta and M. pharaonis homologs.

Run align.py (parallelized by align.sh) to create a codon alignment and compute dN/dS ratios using PAML.

A similar approach was used to find honeybee (Apis mellifera homologs), except OGSv1.0 was used and the evalue was relaxed to 1e-5.

Statistics and plotting

monomorium.R contains all of the statistical analysis.

genome annotation

Round 1

Used training set from Wasmannia and Vollenhovia for first pass. After the run was complete, extracted every prediction that had good transcriptional support

maker2zff  -c 1 -e 1 -o 1

Round 2

SNAPP training

fathom genome.ann genome.dna -gene-stats > gene-stats.log 2>&1
fathom genome.ann genome.dna -validate > validate.log 2>&1
fathom genome.ann genome.dna -categorize 1000 > categorize.log 2>&1
fathom uni.ann uni.dna -export 1000 -plus > uni-plus.log 2>&1
# 3328 genes
mkdir params; cd params
forge ../export.ann ../export.dna > ../forge.log 2>&1
cd ..; hmm-assembler.pl vol params/ > mp.hmm

Augustus training

perl zff2augustus_gbk.pl > genes.gb
etraining --species=generic --stopCodonExcludedFromCDS=false genes.gb 2> train.err
cat train.err | perl -pe 's/.*in sequence (\S+): .*/$1/' > badgenes.lst
# no bad genes

filterGenes.pl badgenes.lst round1.gb > genes.gb

randomSplit.pl genes.gb 200
autoAug.pl --species=ph --genome ../../../ref/Mp.fa --trainingset genes.gb   # 2673987
optimize_augustus.pl --cpus=12  --UTR=on --species=mph genes.gb.train
augustus --species=vem genes.gb.test >train_out.txt
grep -A 22 Evaluation  train_out.txt