STAR alignment

[Ehsan Hajiramezanali]

Hi

I want to use STAR for alignment and I have the following requirements:

1- Can we tune STAR in which for the single end, first n bp only allowed m mismatch?

2- Can we use STAR without the junction alignment?

I wonder if you could help me with these.

Thanks in advance.

Best regards,

Ehsan

[Alexander Dobin]

Hi Ehsan,

>>>1- Can we tune STAR in which for the single end, first n bp only allowed m mismatch?

No, this is not possible.

>>>2- Can we use STAR without the junction alignment?

Yes, with --alignIntronMax 1 at the mapping step and no GTF at the genome generation step.

Cheers

Alex

[Ehsan Hajiramezanali]

Thank you for the kind reply.

>>>1- Can we tune STAR in which for the single end, first n bp only allowed m mismatch?

No, this is not possible.

Can we tune STAR in which for a single end only m mismatch allowed?

[Alexander Dobin]

Hi Ehsan,

if you have single-end reads, you can control the number of mismatches with --outFilterMismatchNmax (10 by default).

Note, that by default STAR will do soft-clipping, and soft-clipped bases are not counted as mismatches.

If you want end-to-end alignments, you would need to use --alignEndsType EndToEnd .

For PE reads, you can only control the total number of mismatches for the whole read (with the same parameter), not for each end separately.

Cheers

Alex

[Ehsan Hajiramezanali]

Thank you very much for the help.

- Is there any way to get the unmapped reads as a separate output in STAR?

- Do you have any suggestion for the number of mismatches for small reads? The default is 10 and it is not proper.

[Ehsan Hajiramezanali]

>>>2- Can we use STAR without the junction alignment?

Yes, with --alignIntronMax 1 at the mapping step and no GTF at the genome generation step.

I created the following example and test --alignIntronMax 1, but I have another conclusion. Based on my example,  by setting --alignIntronMax as 1, STAR will allow junction alignment. Would you please let me know I am in a right track?

~~~

#!/usr/bin/env bash

# vim: set noexpandtab tabstop=2:

tmpdir=$(mktemp -d)

ref=GCACTGTCCGCCAGCCGGTGGATGTGCGGCAACAACATGTCCGCTCCGA

query=${ref:0:14}${ref:20}

ref_fa=$tmpdir/ref.fa

query_fa=$tmpdir/query.fa

cat > "$ref_fa" <<EOF 

>chr

$ref

EOF

cat > "$query_fa" <<EOF 

>seq

$query

EOF

cd "$tmpdir"

mkdir -p genomeDir

source trapdebug.sh

STAR --runMode genomeGenerate --genomeDir genomeDir --genomeFastaFiles "$ref_fa"

outdir1=$tmpdir/outdir1

mkdir -p "$outdir1"

cd "$outdir1"

STAR --genomeDir "$tmpdir/genomeDir" --readFilesIn "$query_fa"

outdir2=$tmpdir/outdir2

mkdir -p "$outdir2"

cd "$outdir2"

STAR --genomeDir "$tmpdir/genomeDir" --readFilesIn "$query_fa" --alignIntronMax 1

function filter {

grep 'Uniquely mapped reads number'

}

diff <(filter < "$outdir1"/Log.final.out) <(filter < "$outdir2"/Log.final.out)

~~~

>>>

~~~

bash> diff <(filter < "$outdir1"/Log.final.out) <(filter < "$outdir2"/Log.final.out)

1c1

<                    Uniquely mapped reads number | 0

---

>                    Uniquely mapped reads number | 1

~~~

[Ehsan Hajiramezanali]

Thanks Alex,

Would you please let me know that two options --outFilterMismatchNoverLmax and --outFilterMismatchNmax have a same meaning?

On Thursday, August 18, 2016 at 11:29:01 AM UTC-5, Alexander Dobin wrote:

[Alexander Dobin]

Hi Ehsan,

to get unmapped reads as fasta(q) file: --outReadsUnmapped Fastx

--outFilterMismatchNmax is the max number of mismatches, while --outFilterMismatchNoverLmax is the max ratio of the number of mismatches to the mapped length

It looks like in your example the read could be mapped with splicing (look for N in the CIGAR), but when splicing is prohibited. no good (unspliced) alignment is found.

Cheers

Alex

[Ehsan Hajiramezanali]

Hi Alex,

Thank you so much to help.

I got my answers about the first two questions, but I am not sure about the third one  

It looks like in your example the read could be mapped with splicing (look for N in the CIGAR), but when splicing is prohibited. no good (unspliced) alignment is found.

Based on my example, I should set  --alignIntronMax (as default) to unspliced alignment and splicing will be prohibited. And --alignIntronMax 1 is for spliced alignment.

Is it correct? 

Actually, I want to do non-spliced alignment and I am searching for this option.

[Alexander Dobin]

Hi Ehsan,

if you set --alignIntronMax 1 , the splicing will be prohibited - i.e. no spliced alignments will be output.

For some reads, this will result in other (non-spliced) alignments reported, often soft-clipped, and for some reads - in complete loss of alignments.

Cheers

Alex