#887 closed task (fixed)

Release export wizard

Reported by: nicklas Owned by: nicklas
Priority: critical Milestone: Reggie v4.5
Component: net.sf.basedb.reggie Keywords:
Cc:

Description (last modified by nicklas)

Implement a wizard for creating all files that should be included in a release. Th wizard should take an item list with raw bioassays as input and produce a lot of files. Unless noted, all files are tab-separated text files.

  • Transcript data (in folder dataTables/transcriptDataTable):
    • tidmatrix.features.txt: Array design features with some annotations. The first line is a header line:
      • id, geneSymbol, refSeq, protAcc, description, chr, entrez.
      • Rows are sorted by internal ID (see comment below for more information)
      • All raw bioassays in the input list must use the same array design.
    • tidmatrix_data.txt: FPKM values for all raw bioassays. Each row represents a feature and each column a raw bioassay.
      • The first line is a header line with raw bioassay names.
      • The first column contains the feature ID.
      • Same order of rows as the tidmatrix.features.txt.
    • tidmatrix_FPKM_conf_hi.txt, tidmatrix_FPKM_conf_lo.txt, tidmatrix_FPKM_status.txt: More data files similar to the tidmatrix_data.txt file but with the FPKM_conf_hi, FPKM_conf_lo and FPKM_status values.
  • Gene data (in folder dataTables/geneDataTable):
    • genematrix_data.txt: Sum of FPKM values per gene symbol.
      • The first line is a header line with raw bioassay names.
      • The first column is the gene symbol (in no particular order).
    • is.NM.gene.txt: TRUE/FALSE flag for each gene indicating if the refSeq ID starts with NM_ or not.
      • No header line.
      • Rows must be in the same order as in genematrix_data.txt.
      • First column is the line number (in this file, add +1 for getting the line number in genematrix_data.txt).
      • Second column is TRUE or FALSE.
      • Third column is the gene symbol.
  • Cohort data (in folder cohortTables): A set of tab-separated files with data for each raw bioassay and the parent items it is derived from. Each file starts with a header line. Each row contains data for one raw bioassay. The first column (rba) is always the name of the raw bioassay. Columns ending with .A. are annotation columns. Date values are formatted as YYYY-MM-DD unless otherwise noted.
    • cohortRawbioassay.txt: Data from the raw bioassay level. Columns:
      • ID: Internal ID in BASE
      • Name: Name of raw bioassay
      • Platform: Name of platform (Sequencing)
      • Raw.data.type: Name of raw data type (cufflinks)
      • Has.data: Flag indicating if there is raw data for this raw bioassay or not (TRUE/FALSE)
      • Db.spots: Number of raw data entries
      • Array.design: Name of the array design
      • Software: Name of the software used to generate the raw data
      • Import.date: Date the raw data was created
      • AnalysisResult..A.
      • DataFilesFolder..A.
      • FPKM.tracking.file..F.: Path to the isoforms.fpkm_tracking file in the BASE file system
    • cohortAligned.txt: Data from the AlignedSequences parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (AlignedSequences)
      • Software: Name of the software used for alignment
      • Registered: Date the item was registered in BASE
      • AnalysisResult..A.
      • DataFilesFolder..A.
      • ALIGNED_PAIRS..A.
      • READ_PAIRS_EXAMINED..A.
      • READ_PAIR_DUPLICATES..A.
      • FRACTION_DUPLICATION..A.
      • FragmentSizeAvg..A.
      • FragmentSizeStdev..A.
    • cohortMasked.txt: Data from the MaskedSequences parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (MaskedSequences)
      • Software: Name of the software used for masking
      • Registered: Date the item was registered in BASE
      • PM_READS..A.
    • cohortMerged.txt: Data from the MergedSequences parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (MergedSequences)
      • Physical.bioassays: Name of the physical bioassay (flow cell) used for sequencing. Comma-separated list if there is more than one.
      • Software: Name of the software used for merging
      • Registered: Date the item was registered in BASE
      • AnalysisResult..A.
      • DataFilesFolder..A.
      • READS..A.
      • PF_READS..A.
      • ADAPTER_READS..A.
      • PT_READS..A.
      • FragmentSizeAvg..A.
      • FragmentSizeStdev..A.
    • cohortSequencing.txt: Data from the SequencingRun parent item. Columns:
      • TODO
    • cohortLibrary.txt: Data from the Library parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Library)
      • Protocol: Name of the library preparation protocol
      • Created: Date the library was created
      • Tag: Name of the barcode used by the library
      • Bioplate: Name of the library plate
      • Biowell.row: Row coordinate on the library plate (A-H)
      • Biowell.column: Column coordinate on the library plate (1-12)
      • QubitConc..A.
    • cohortRNA.txt: Data from the RNA/RNAQC parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (RNA)
      • Original.quantity..µg.: µg RNA that was extracted
      • NDConc..A.
      • ND260by230..A.
      • ND260by280..A.
      • QiacubeDate..A.
      • QiacubeRunNo..A.
      • QiacubePosition..A.
      • RNAQC_last: RIN/RQS value from the latest quality control
    • cohortLysate.txt: Data from the Lysate parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Lysate)
      • Created: Date the lysate was created
      • Original.quantity..µg.: µg Lysate that was extracted
      • Parent.items: Name and used quantity from the specimen
      • MultPieces..A.
      • PartitionDate..A.
    • cohortSample.txt: Data from the Specimen parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Specimen)
      • Original.quantity..µg.: Weight of sample received
      • Created: Date the sample was created (operation date)
      • ArrivalDate..A.
      • BiopsyType..A.
      • SpecimenType..A.
      • Laterality..A.
      • NofDeliveredTubes..A.
      • NofPieces..A.
      • OperatorDeliveryComment..A.
      • OperatorPartitionComment..A.
      • SamplingDateTime..A.
      • RNALaterDateTime..A.
      • LinkedSpecimen..A.
    • cohortCase.txt: Data from the Case parent item (except INCA data). Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Case)
      • Consent..A.
      • ConsentDate..A.
      • Laterality..A.
    • cohortPatient.txt: Data from the Patient parent item. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Patient)
      • Gender..A.
      • Samples: Comma-separated list with all child item names (cases and blood)
    • cohortStained.txt: Data from the Stained parent item. To find the correct item we need to descend from the specimen level (Specimen -> Histology -> Stained) and select the one which has GoodStain=TRUE. Columns:
      • ID: Internal ID in BASE
      • Name: Name of item
      • Type: Type of item (Case)
      • Created
      • Registered
      • Bioplate
      • Biowell.row
      • Biowell.column
      • GoodStain..A.
      • ScoreComplete..A.
      • ScoreInvasiveCancer..A.
      • ScoreInsituCancer..A.
      • ScoreLymphocytes..A.
      • ScoreStroma..A.
      • ScoreFat..A.
      • ScoreNormal..A.
    • cohortINCA.txt: Data from parent items (eg. Case) that have been imported from the INCA registry. Columns:
      • IncaExportDate..A.
      • All other annotation types in the INCA category. The INCA_ prefix is removed. No ..A. is added in the header.
    • cohortSummaryTable.txt: A single table collecting some of the most useful information from the other tables. See the summary columns text file. The order of the columns in the file is not correct. They should have the same order as in the main data files.
  • Subtype data (in folder cohortTables/subtypeTables): Information generated by the R report scripts. We do not currently store this information in BASE, so it needs to be discussed how this should be done. The report plug-in could for example import the data from the R scripts as annotations.
  • README files
    • TODO

Attachments (4)

summary_columns.txt (1.3 KB) - added by nicklas 18 months ago.
Columns in the cohortSummary.txt
join-order-by-external-id.png (8.5 KB) - added by nicklas 18 months ago.
Sort by external reporter id
join-order-by-internal-id.png (8.8 KB) - added by nicklas 18 months ago.
Join and sort by internal reporter id
no-join-order-by-internal-id.png (2.2 KB) - added by nicklas 18 months ago.
No join sort by internal reporter id

Download all attachments as: .zip

Change History (52)

comment:1 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:2 Changed 19 months ago by nicklas

  • Description modified (diff)

For efficient calculations it is desirable to process the data gene symbol by gene symbol. Thus, the data must come sorted in gene symbol order.

comment:3 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:4 Changed 19 months ago by nicklas

  • Status changed from new to assigned

comment:5 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:6 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:7 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:8 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:9 Changed 19 months ago by nicklas

  • Milestone changed from Reggie v4.x to Reggie v4.5

comment:10 Changed 19 months ago by nicklas

(In [3932]) References #887: Release export wizard

Created the exporter plug-in (ReleaseExporterPlugin). This is a front-end for interacting with the user. The plug-in is installed in the "Item list" toolbar and asks for an output directory and if existing files should be overwritten or not.

The actual export is going to be implemented in the ReleaseExporter class. It is currently only able to collect the parameters and make some checks on the rawbioassays/array design.

comment:11 Changed 19 months ago by nicklas

(In [3933]) References #887: Release export wizard

Implemented exporter for the array design. This creates the tidmatrix.features.txt and is.NM.gene.txt files.

comment:12 Changed 19 months ago by nicklas

(In [3934]) References #887: Release export wizard

Implemented exporter for th transcript data. Eg. all remaining files in the dataTables/transcriptDataTable directory.

Note that the current implementation only export data from the first raw bioassay. BASE 3.9 is needed to be able to export from all raw bioassays.

See http://base.thep.lu.se/ticket/2004

comment:13 Changed 19 months ago by nicklas

(In [3935]) References #887: Release export wizard

Started with the cohort exporters. The CohortItem is used for loading all information related to a single raw bioassays. It is used and then discarded in smaller batches to let garbage collection clean up memory.

The CohortWriter is an extension to the ReleaseWriter for writing cohort data. Each cohort data file need a subclass. The RawBioAssayWriter is the first such writer that write the raw bioassay data.

comment:14 Changed 19 months ago by nicklas

(In [3936]) References #887: Release export wizard

Added exporters for aligned, masked and merged data tables.

comment:15 Changed 19 months ago by nicklas

(In [3937]) References #887: Release export wizard

Added exporters for library and RNA data tables.

comment:16 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:17 Changed 19 months ago by nicklas

(In [3938]) References #887: Release export wizard

Set character set (UTF-8) and MIME type (text/plain) on exported files.

comment:18 Changed 19 months ago by nicklas

(In [3939]) References #887: Release export wizard

Added exporters for lysate and specimen data tables.

comment:19 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:20 Changed 19 months ago by nicklas

(In [3940]) References #887: Release export wizard

Added exporters for case and patient data tables.

comment:21 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:22 Changed 19 months ago by nicklas

  • Description modified (diff)

comment:23 Changed 19 months ago by nicklas

(In [3941]) References #887: Release export wizard

Added exporter for the Stained data table.

comment:24 Changed 19 months ago by nicklas

(In [3942]) References #887: Release export wizard

Added exporter for the INCA data table.

comment:25 Changed 18 months ago by nicklas

(In [3943]) References #887: Release export wizard

Started to implement the summary exporter. It has been implemented to copy columns from the other exporters. An extra complication is that the headers in the summary are different from the headers in the main data files. This is handled by adding the CohortHeader class which can hold two different headers. One for the main file and one for the summary.

So far, only the patient and case information is copied to the summary.

comment:26 Changed 18 months ago by nicklas

  • Description modified (diff)

Changed 18 months ago by nicklas

Columns in the cohortSummary.txt

comment:27 Changed 18 months ago by nicklas

(In [3944]) References #887: Release export wizard

Added specimen and stained information to the summary.

comment:28 Changed 18 months ago by nicklas

(In [3945]) References #887: Release export wizard

Added lysate, RNA and library information to the summary.

comment:29 Changed 18 months ago by nicklas

  • Description modified (diff)

comment:30 Changed 18 months ago by nicklas

(In [3946]) References #887: Release export wizard

Added merged, masked, aligned and raw bioassay information to the summary.

comment:31 Changed 18 months ago by nicklas

(In [3947]) References #887: Release export wizard

Added INCA information to the summary.

comment:32 Changed 18 months ago by nicklas

(In [3948]) References #887: Release export wizard

Implemented exporter for the genematrix_data.txt. There are (at least) two problems with the current approach:

  • The database will probably choke on the "ORDER BY symbol" clause when used with a large set since this is not an indexed column. The query analyzer in Postgres shows that the join is done first and then the sort (for a data set with 1000 raw bioassays this is 100M rows to sort).

  • The generated file is not compatible with the row numbering in is.NM.gene.txt since the raw bioassays doesn't contain data for all features.

Changed 18 months ago by nicklas

Sort by external reporter id

Changed 18 months ago by nicklas

Join and sort by internal reporter id

Changed 18 months ago by nicklas

No join sort by internal reporter id

comment:33 Changed 18 months ago by nicklas

I have made some tests with different strategies for loading the raw data and producing the tidmatrix.* files. We want to avoid keeping too much data in memory at the same time which means that we must process data ordered by feature. Basically, we can choose to sort the data either by external id or internal id. The currently released files are sorted by external id.

I checked a few different queries in the PostgreSQL query planner.

Sort by external reporter id

select r.external_id, c.fpkm 
from "RawDataCufflinks" c 
inner join "Reporters" r on r.id=c.reporter_id
order by r.external_id

Sort by external reporter id

Here raw data is retrieved via a sequential scan which is ok since we are going to need all in any case. The hash join to reporters is not a problem. The last sort step may turn out to be expensive though since it needs to sort 100M+ rows.

Sort by internal reporter id (keeping the join)

select r.external_id, c.fpkm 
from "RawDataCufflinks" c 
inner join "Reporters" r on r.id=c.reporter_id
order by c.reporter_id

Join and sort by internal reporter id

Here the raw data is retrieved in the desired order by using an index scan (FKA74...). The sort on reporter and the merge join should not too expensive.

Sort by internal id (without joining the reporters table)

select c.reporter_id, c.fpkm 
from "RawDataCufflinks" c
order by c.reporter_id

No join sort by internal reporter id

This seems like the best alternative of them all. Only the sequential index scan is needed. We need to keep some reporter information in memory so we can get the external id from the internal id but that is not using very much memory.

comment:34 Changed 18 months ago by nicklas

  • Description modified (diff)

comment:35 Changed 18 months ago by nicklas

(In [3953]) References #887: Release export wizard

Re-factored the array design export and the transcript data export.

The array design information is now only pre-loaded into memory in the first step (FeatureInfo).

The transcript data export was changed to sort by internal reporter id instead of the external reporter id (see comment:33).

The array design file (tidmatrix.features.txt) is now written (by FeatureWriter) at the same time as the transcript data files and always in the same order and with the same number of rows.

comment:36 Changed 18 months ago by nicklas

(In [3954]) References #887: Release export wizard

Added GeneSymbolInfo for keeping track of information related to a gene symbol. This information is used by the IsNmGeneWriter to create the is.NM.gene.txt file.

Generation of the genematrix_data.txt has temporarily been disabled. Need to figure out a way to create this at the same time as the other data matrix files.

comment:37 Changed 18 months ago by nicklas

  • Description modified (diff)

comment:38 Changed 18 months ago by nicklas

(In [3955]) References #887: Release export wizard

Re-factored the code for creating the genematrix_data.txt file. The GeneSummaryWriter is a special writer implementation in the first stage simply collect the FPKM values and store them in memory. After the entire data set has been processed the summarized data is writted to the file.

The drawback with this implementation is that it may use too much memory. This should be investigate more.

comment:39 Changed 18 months ago by nicklas

(In [3956]) References #887: Release export wizard

Changed the GeneSummaryWriter to use float[] instead of Object[] which lowers memory usage a lot. The drawback is that we need to copy values back to an Object[] before writing but that seems to be fast enough and we can re-use the same instance.

Added a progress reporter when writing the gene data file.

comment:40 Changed 18 months ago by nicklas

(In [3957]) References #887: Release export wizard

Implemented a new strategy for writing the genematrix_data.txt file.

When loading the array design information we count the number of transcripts with the same gene symbol and keep that number in GeneSymbolInfo.

Then when exporting the data we also count the number of transcripts and when we have reached the final count we know that it is safe to write the summed FPKM values to the file and we can release the temporary data from memory. Gene symbols with only one transcript are a special case (almost 50%) which are written immediately to the file.

This changes the order of the gene symbols and the same change must be applied in IsNmGeneWriter to make sure the order matches.

Finally, since the raw data is lacking a few transcripts there are a few (14) that never reaches the final count. They are handled by in the close() method.

Checking some statistics for this method showed that the maximum numbed of gene symbols we need to keep in memory at the same time was less than 2000. Memory usage seemed to stay well below 1GB also when simlulating an export for 10000 raw bioassays.

comment:41 Changed 18 months ago by nicklas

(In [3958]) References #887: Release export wizard

Added exporter for the sequencing data table.

comment:42 Changed 18 months ago by nicklas

(In [3959]) References #887: Release export wizard

Parent.items column in the lysate export changed to ParentQuantityUsed..µg.. Only the actual used quantity is exported, not the name of the parent item (which can be found in the cohortSample.txt table).

Fixed formatting of PartitionDate values.

comment:43 Changed 18 months ago by nicklas

(In [3961]) References #887: Release export wizard

The exporter now uses the new method for loading raw data for all raw bioassays in a single query.

Removed debug code.

comment:44 Changed 18 months ago by nicklas

(In [3962]) References #887: Release export wizard

Progress reporter for the cohort step should go from 75-99%.

comment:45 Changed 18 months ago by nicklas

(In [3971]) References #887: Release export wizard

Improved progress reporting.

comment:46 Changed 18 months ago by nicklas

(In [3978]) References #887: Release export wizard

Progress report every 10 items when exporting cohort data. The 100 item interval took just a bit to long to be comfortable.

comment:47 Changed 18 months ago by nicklas

(In [3979]) References #887: Release export wizard

Added checks for missing items so we can get better error messages instead of NullPointerException.

Verify that the names of the raw bioassays look like SCANB names.

Make a pre-check in the job configuration phase to catch some errors before starting the job.

comment:48 Changed 18 months ago by nicklas

  • Resolution set to fixed
  • Status changed from assigned to closed
Note: See TracTickets for help on using tickets.