Submission of IARC gene inference data to NCBI

Submission of IARC gene inference data to NCBI#

General outline#

IARC submission currently follows a “INSDC first” approach, means that all sequence data related to the reported inference is REQUIRED to be properly deposited in a general purpose sequence repository before it is reviewed by IARC. The submitter needs to complete the initial steps of submission to one of the INSDC repositories. Upon submission to IARC, some of this data will be pulled in from NCBI (TODO: What kind of data can we actually pull down from INSDC?)

The aim of this procedure is to ensure that inferences reviewed by IARC are public and will remain available in the long run. It is however explicitly not the aim to provide data that deterministically will yield the same inference results.

Deposition of inferred gene data at NCBI#

At the end of the deposition process there should be three types of records present at NCBI:

  1. A single record containing the final and full-length inferred sequence. The record is deposited in one of the following:

    • Genbank: All inferences that have been performed on the submitters own data CAN be submitted as [???] to Genbank. Note that Genbank typically only holds data that has a physical correlate which is not necessarily true for inferred sequences. Nevertheless NCBI currently accepts this as a kind of consensus building if it is performed on your own data. The Genbank record MUST link to the select set record (see 3.) via the DBLINK/DR field. Genbank records will be publicly available independent of other publications. Note that the for Genbank, the DBLINK field does not appear to be available through the BankIt submission interface. You can use Tbl2asn and Sequin, and edit the DBLINK field manually (as “Sequence Read Archive” is not one of the options on the template creation page. A sample Genbank deposit can be found under accession MK321694.

    • TPA (Third-party annotation): A segment of Genbank dedicated inferences. Also the TPA record MUST link to the select set record (see 3.) via the DBLINK/DR field. Note that in contrast to Genbank, TPA does REQUIRE a peer-reviewed publication describing the details of the inference process before the record will be made publicly available. A sample TPA deposit can be found under accession BK01573.

    The format for both record types the Genbank format (link) with a standardized feature table (FT). Note that your initial submission MUST NOT contain any potential name for the gene as this will be assigned by IARC later on.

    TODO: Is there any metadata that should be provided into the GB record?

  2. One or multiple SRA records containing all raw reads of the the respective sequencing run. Note that if you are performing inference using third-party data, these records MUST be submitted by the original owner of the data. These record type will typically be present before the other. The metadata annotation of the records SHOULD be MiAIRR compliant [Rubelt et al.].

  3. One or multiple SRA records containing the select set of reads from (2). The aim of these records is to document the number, quality, coverage and diversity of the reads in a dataset that _potentially_ support the inference. This means that the select set SHOULD be a superset of the reads that support the inference. It is NOT REQUIRED that inference tools deterministically return the inferred allele upon being fed with the select set. Generation of the select set from the complete set is described below. When submitting the select set to SRA the metadata context, i.e. the original links to project, sample and (if possible) experiment) SHOULD be maintained. Reads originating from multiple subjects or samples MUST NOT be pooled into a single new entry. The new record SHOULD be titled “Reads from <original_run_accession> supporting inference of Homo sapiens immunoglobulin heavy chain variable gene” and contain a design description, e.g., “Experimental workflow as described in original SRA/ENA record [<run_accession>]. Gene inference was performed using <software+version+parameters>. The reported reads were selected based on <selection_criteria>.”

NOTE: It is reasonably likely, in the short term, that you will encounter questions from the SRA/ENA/Genbank staff about the nature of these deposits. If so, you can respond that they are made as part of a community effort to document novel alleles with an emphasis on transparency in data provenance. You can link to the IARC page and note that we worked together with IMGT and Genbank/TPA staff in designing this procedure.

Generating the select set#

Below is the current procedure describing how to generate a select set using general purpose tools. This procedure was designed in a rather generic fashion so that it is easy to implement and does NOT REQUIRE inference tools to provide their own mechanisms. Note that it is currently assumed that the procedure is not fully deterministic, i.e. the select set cannot simply be generated using the complete read data and the inferred sequence, as there are additional filter criteria that apply. In addition the select set SHOULD not be subject to any modifications that are not listed below. This includes UMI-based consensus building or other aggregation steps that are not fully transparent to a third-party.

  1. Assemble paired-end reads. The two reads MUST overlap. Recommended tool: PandaSeq

  2. Perform PHRED filtering that is equivalent to the one performed by inference pipeline. Recommended tool: Immcantation suite

  3. Perform a blastn search using the data from (2.) as query and bp 1-312 of the inferred gene as reference library. Require matches to be full-length and >99.6% ID. Record all matching read ID. Recommended tool: NCBI BLAST

  4. Select the reads with the read ID found in (3.) from the original unmerged FASTQs. Note that each select set MUST be derived from a single donor and sample. Recommended tool: Christian’s cryptic extractor script

  5. Submit the select set to SRA