BaseSpace ChIP-Seq output conversion to bioset

We currently do not support the curation of public ChIP-Seq analyses as a service for Correlation Engine. This document outlines the general process by which we would transform the output of the BaseSpace ChIP-Seq application into a format suitable for upload as a bioset. There are two data files generated by the BaseSpace Sequence Hub ChIP-Seq application that can be used for bioset creation for upload into Correlation Engine.

If the experiment is characterizing the genome occupancy by modified histones, the data ChIP-Seq_peaks.xls file is used. If the study was of transcription factor binding or RNA polymerase binding, the data from the ChIP-Seq_peaks.narrowPeak file is as well as settings information provided in the ChIP-Seq_peaks.xls file.

Histone binding

The ChIP-Seq_peaks.xls contains a header area offset with hashes (#) with settings values and other metadata around processing and data section with the following column headers:

Original header	Renamed header
chr	chr
start	Start
end	End
length
abs_summit
pileup
=-LOG10(pvalue)	p-value
fold_enrichment	Enrichment fold
=-LOG10(qvalue)	q-value
name

The renamed header column indicates which data should be used and how the header names should appear in the bioset before upload. The values in the p-value and q-value columns should be multiplied by -1 and unlogged to generate the original p- and q-values.

The data columns should be pulled out, transformed appropriately and added to tab-delimited text file with a header describing the settings used in processing. For example:

Bioset summary = Mammary epithelial cell genome occupancy by MYC ChIP-Seq
Bioset description = Mammary Epithelial cells expressing HA-tagged MYC + MYC ChIP _vs_ input ChIP-Seq
Organism = Homo sapiens
Data pre-processing = FASTQ files were downloaded from Sequence Read Archive (SRA). No other preprocessing was performed.
Analysis summary = Identification of chromosome regions significantly enriched by chromatin immunoprecipitation (ChIP). Sequence reads from each IP experiment are analyzed using the BaseSpace Sequence Hub ChIPSeq application version 1.0.2 https://basespace.illumina.com/apps/2561559/ChIPSeq?preferredversion. Sequence reads are aligned to the Homo sapiens reference genome (NCBI37/hg19) using BWA and processed using MACS2 2.1.0.. Parameter settings used = [Fragment size = 200, Q-value cutoff = 0.01, Fold enrichment = 10,30, Sample normalization = Down Sample]. Regions, defined by the chromosome start and end positions, are ranked by fold change as determined by comparison to input DNA or IgG ChIP, as provided. For transcription factor binding studies, narrow peak calls are reported. For histone modifications, broad peak calls are reported.
Test group = Immortalized_mammary_epithelial_cell_DNA_expressing_HA-tagged_MYC_+_MYC_chIP
(total replicates = 1)
Control group = Immortalized_mammary_epithelial_cell_DNA_expressing_HA-tagged_MYC_-_input (total replicates = 1)
Platform(s) = Illumina iGenome UCSC, hg19, March 6, 2013 RefSeq
Internal ID - GSE66252_1
chr	start	end	Enrichment fold	q-value	p-value
chr7	4681591	4681891	34.34069	4.92E-97	3.97E-105
chr17	61851121	61851429	28.33336	2.10E-96	1.90E-104
chr17	75084699	75084869	26.33382	2.71E-85	7.89E-93
chr1	1.59E+08	1.59E+08	21.00002	4.24E-81	1.86E-88
chr2	2.17E+08	2.17E+08	36.57338	5.76E-81	2.63E-88

Q-value cut-offs are determined in the processing set-up, usually with a default of q-value = 0.01. No additional cut-offs need be applied before upload.

Transcription factor binding

Transcription factor protein-DNA binding experiments should report calls provided in the ChIP-Seq_peaks.narrowPeak file, which looks like the following:

Note the lack of column headers or any processing related information. The content of the columns is as follows:

1. Chromosome number

2. Start

3. End

4. Name

5. Integer score for display

6. Strand

7. Fold change

8. –log10pvalue

9. –log10qvalue

10. Relative summit position to peak start

Columns 1, 2, 3, 7, 8, and 9 should be brought together in a new table with the following headers applied:

1. chr

2. start

3. end

4. Enrichment fold

5. p-value

6. q-value

The p-values and q-values should be multiplied by -1 and unlogged.

Header information should be added as for histone biosets, drawing the settings information from the ChIP-Seq_peaks.xls file. The genome build can be read from sample level statistics files if needed, located at <ChIP-Seq>/Statistics/<SRR#######_S#.summary.csv>.

Correlation Engine annotates genomes upon upload using one particular genome build per supported species. Lift-over tools are available that allow the user to convert a subset of other builds to the Correlation Engine standard. The following table outlines what lift-over tools are available and the target build.

Species	Build	Lift-over to
Human	hg19, Genome Reference Consortium GRCh37 hg18, NCBI Build 36.1 hg17, NCBI Build 35 hg16, NCBI Build 34 hg15, NCBI Build 33	Hg18
hg19, Genome Reference Consortium GRCh37
hg18, NCBI Build 36.1
hg17, NCBI Build 35
hg16, NCBI Build 34
hg15, NCBI Build 33
hg19, Genome Reference Consortium GRCh37
hg18, NCBI Build 36.1
hg17, NCBI Build 35
hg16, NCBI Build 34
hg15, NCBI Build 33
Mouse	mm9, NCBI Build 37 mm8, NCBI Build 36 mm7, NCBI Build 35 mm6, NCBI Build 34 mm5, NCBI Build 33	Mm9
mm9, NCBI Build 37
mm8, NCBI Build 36
mm7, NCBI Build 35
mm6, NCBI Build 34
mm5, NCBI Build 33
mm9, NCBI Build 37
mm8, NCBI Build 36
mm7, NCBI Build 35
mm6, NCBI Build 34
mm5, NCBI Build 33
Rat	Rn4, NCBI RGSC v3.4	Rn4
Worm	ce6, NCBI WS190 ce4, NCBI WS170 ce2, NCBI WS120	Ce6
ce6, NCBI WS190
ce4, NCBI WS170
ce2, NCBI WS120
ce6, NCBI WS190
ce4, NCBI WS170
ce2, NCBI WS120
Fly	dm3, NCBI v5.2 dm2, NCBI v4 dm1 NCBI v3	Dm3
dm3, NCBI v5.2
dm2, NCBI v4
dm1 NCBI v3
dm3, NCBI v5.2
dm2, NCBI v4
dm1 NCBI v3
Saccharomyces cerevisiae	sacCer2 sacCer1	sacCer2
sacCer2
sacCer1
sacCer2
sacCer1
Monkey	rheMac2	rheMac2
Chicken	galGal3	galGal3
Chimp	panTro2	panTro2
Cow	bosTau4	bosTau4
Fish	danRer6	danRer6
Dog	canFam2	canFam2

Upon upload vie the Correlation Engine user interface, select the appropriate data type and species and whether the experiment type was such that multiple samples were grouped together as aggregates for analysis, or whether just single samples were used, i.e., 1 IP and 1 input sample, for the comparison.

After choosing the biosets files and all the required information is recognized in the next page, select the Enrichment fold column for ranking, and Rank, descending for how the bioset should be ranked. Lastly on this page select the original genome build that was used for generating the data. For example, if hg19 was used for alignments, select hg19 to lift-over coordinates from that build to hg18. The user has an opportunity to examine any unrecognized rows or rows dropped by the build conversion.

The last step before commencing upload is to apply tags to the biosets. The standard types of tags are BIOSOURCE, TISSUE, the GENE that is the antibody target, the GENEMODE “ChIP antibody target” and an appropriate biodesign of Normal vs. normal, Disease vs. disease, Treatment vs. treatment. We do not normally ingest protein-DNA studies of designs other than ChIP _vs_ input or ChIP _vs_ Ig.