TSS assembly pipeline for Dm_EPDnew_005


This document provides a technical description of the transcription start site assembly pipeline that was used to generate EPDnew version 005 for D. melanogaster.

Source Data

Promoter collection:

Name Genome Assembly Promoters Genes PMID Access data
ENSEMBL86 Aug 2014 BDGP Rel6 + ISO1 MT/dm6 18409 13660 19420058 SOURCE DOC DATA

Experimental data:

Name Type Samples Tags PMID Access data
MachiBase OligoCap 7 24,984,353 18842623 SOURCE DOC DATA
Hoskins et al., 2012 CAGE 1 17,979,809 21177961 SOURCE DOC DATA
modENCODE CAGE 49 596,317,845 24985915 SOURCE DOC DATA
Ni et al., 2010 CAGE 2 27,173,616 20495556 SOURCE DOC DATA
Schor et al., 2017 CAGE 316 2,536,326,545 28191888 SOURCE DOC DATA

Assembly pipeline overview

Description of procedures and intermediate data files

1. Biomart Download

Data was downloaded from BioMart selecting the following attributes:

  1. Ensembl Gene ID
  2. Ensembl Transcript ID
  3. Chromosome Name
  4. Strand
  5. Transcript Start (bp)
  6. Transcript End (bp)
  7. Gene Start (bp)
  8. Gene End (bp)
  9. Status (transcript)
  10. Status (gene)
  11. Associated Gene Name

Then, transcrips have been filtered according to the following rules:

  1. Transcripts of protein coding genes only
  2. Transcript length > 0 [Transcript Start different from Transcript End]
  3. Transcript lies on full chromosomes
  4. Gene must have a 5' UTR [Transcript Start different from Gene Start]
  5. Genes must be annotated [Associated Gene Name present]
  6. Gene and transcripts status known

Gene names were taken from the field 'Associated Gene Name'. Since the EPD format doesn't allow gene names longer than 18 characters, we checked whether the names repsected this limitation.

Transcripts with the same TSS position were merged under a common ID. As a conseguence of this, from the 23850 transcrips originally present in the ENSEMBL database, 5953 were merged, leaving 17897 uniquely mapped promoters in the input list.

2. EMBL TSS collection

The ENSEMBL TSS collection is stored as a tab-deliminated text file conforming to the SGA format under the name:

  • Dm_ENSEMBL70.sga

The six field contain the following kinds of information:

  • NCBI/RefSeq chromosome id
  • "TSS"
  • position
  • strand ("+" or "-")
  • "1"
  • gene name.

Note that the second and forth fields are invariant.

3. Data import from MachiBase

MachiBase data were generated with the oligo-capping technology. The source data were downloaded from:

According to the readme file included in the tar archive, the 5' end tags were mapped to the Drosophila genome using BLAT as alignment tool allowing for up to three mismatches.

4. oligocap tags

The compressed version of this file is available from the MGA archive (see above) under the name:


5. Data import from Genome Research

Mapped sequence tags were extracted from Supplementary Data File 1 available from Genome Research at:

The downloaded source file is in SAM format and has been generated with the tag mapping program StatMap as described in the article cyted above. We extracted all tags with mapping quality scores greater or equal to 30.

6. CAGE tags

The compressed version of this file is available from our ftp site (see above link) with the name:


7. Data import from SRA

BAM files for the SRA serie SRP001602 and SRX018832 were downloaded from SRA site and converted into SGA file using in house software.

8. Data import from ArrayExpress

FASTQ files for the ArrayExpress serie E-MTAB-4787 were downloaded from ENA site and mapped to the genome with bowtie. File convertion into SGA format was performed using standard software. A ditailed description of the procedures involved can be found in the documentation.

9. TSS validation and shifting

Each sample in the collection (mRNA peaks and ENSEMBL TSS) was then processed in a pipeline aiming at validating transcription start sites with mRNA peaks. An Ensembl TSS was experimentally confirmed if a CAGE peak lied in a window of 200 bp around it and if it had a maximum high of at least 3 tags. The validated TSS was then shifted to the nearest base with the higher tag density.

10. Promoter collection for each sample

Each sample in the dataset was used to generate a separate promoter collection. Potentially, the same transcript could be validated by multiple samples and it could have different start sites in different samples. To avoid redundancy, the individual collections were used as input for an additional step in the analysis (Assembly pipeline part B).

11. Quality checks for each sample-specific promoter collection

The quality of each sample-specific promoter colection was evaluated studing the density distribution of known core promoter elements at the expected positions. A TATA-box score was evaluated as the density distribution of the lement at position -29 from the TSS and an Initiator score as the density disribution at the TSS. Samples with very low scores (outliers) were discarded from the pool.

12. Merging collections and second TSS selection

All promoter collections were merged into a unique file and further analysed with the exception of samples from Schore et al., 2017. Since the promoter of a transcript was mantained in the list only if validated by at least two samples and the position validated by the largest number of samples selected as EPDnew promoter, there would have been an inbalance in TSS selection given the large number of samples in Schore et al., 2017 coming from only 3 developmental stages. For this reason, in order to reduce the inpact of these developmental stages in TSS selection we first generate a developmetal-stage-specific TSS collection and then merged these with the remaining samples from the other sources. A further selection was applyed to choose the promoter position validaed by the largest number of samples.

13. Filtering

Transcription Start Sites that mapped closed to other TSS that belonged to the same gene (200 bp window) were merged into a unique promoter following the same rule: the promoter that was validated by the higher number of samples was kept.

14. EPDnew collection

The experimentally validated promoter were stored in the EPDnew database that can be downloaded from our ftp site. Scientist are wellcome to use our other tools ChIP-Seq (for correlation analysis) and SSA (for motifs analysis around promoters) to analyse EPDnew database.

Last update May 2017