Reece Hart, Ph.D. Reece Hart, Ph.D. [email protected][email protected]Genentech Genentech 2014-10-16 2014-10-16 The Clinical Significance of Transcript The Clinical Significance of Transcript Alignment Discrepancies Alignment Discrepancies … … and tools to help you deal with them. and tools to help you deal with them. Available on SlideShare (hp://www.slideshare.net/reecehart)
28
Embed
Clinical significance of transcript alignment discrepancies gne - 20141016
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
The Clinical Significance of Transcript The Clinical Significance of Transcript Alignment DiscrepanciesAlignment Discrepancies… … and tools to help you deal with them.and tools to help you deal with them.
Available on SlideShare (http://www.slideshare.net/reecehart)
2 / 28
The fidelity of transcript-genome mapping matters.The fidelity of transcript-genome mapping matters.
Variants are identified and computed on in genome coordinates
Variants are analyzed and communicated using
transcript coordinatesgenome totranscript(g. to c.)
transcriptto genome
(c. to g.)
3 / 28
Motivation 1: Discordant exon coordinatesMotivation 1: Discordant exon coordinatesNCBI and UCSC report different coordinates for CARD9, NM_052813.3, exon 12NCBI and UCSC report different coordinates for CARD9, NM_052813.3, exon 12
UCSC(BLAT)
NCBI(Splign)
Consequences:1. An assay that targets the wrong genomic region will generate uninformative sequence data.2. A genomic variant will be interpreted as exonic when it is intronic, or vice versa.
exon 12displaced 322 nt
4 / 28
Motivation 2: indels confound mappingMotivation 2: indels confound mappingNM_006158.3 (NEFL) contains indel in CDSNM_006158.3 (NEFL) contains indel in CDS
Deletion justified differently!
5 / 28
Motivation 3: Data management challengesMotivation 3: Data management challenges
➢ Mutable data (!)➢ Sporadic failures➢ Inconsistent data from a single source➢ Inconsistent data across sources➢ Opaque and implicit data definitions➢ Historical alignment data not available
Motivation 4: Use Ensembl for Variant Effect PredictionMotivation 4: Use Ensembl for Variant Effect Prediction
RefAgreeDo transcript and genome sequences agree?
Transcript EquivalenceWhich RefSeq and Ensembl transcripts are equivalent?
RefSeq(NM)
Ensembl(ENST)
Genome(GRCh37)
➊ SNV
➌
➋ Indel
➍ Historical Transcripts UCSC (NM)LRG, BIC, …
7 / 28
Garla, V., Kong, Y., Szpakowski, S., & Krauthammer, M. (2011).MU2A--reconciling the genome and transcriptome to determine the effects of base substitutions.Bioinformatics (Oxford, England), 27(3), 416-8. doi:10.1093/bioinformatics/btq658
8 / 28
Challenges and Solutions in Transcript ManagementChallenges and Solutions in Transcript Management
➢ Biological● Alternative splicing● Paralogs● Natural polymorphisms● Alternative references
““RefAgree” Statistics by Protein Coding TranscriptRefAgree” Statistics by Protein Coding TranscriptSequence concordance between RefSeq and GRCh37 primary assemblySequence concordance between RefSeq and GRCh37 primary assembly
c.f. Garla V, et al. Bioinformatics 27(3): 416–8 (2010).
34531 NM transcripts (Jan 2014)760 0.2% with length discrepancies
3481 10% with substitutions321 0.9% with deletions255 0.7% with insertions
➊➋
17 / 28
Exon structures have unique fingerprintsExon structures have unique fingerprintsIdentifying ENST-NM equivalences with fingerprintsIdentifying ENST-NM equivalences with fingerprints
=> select N.hgnc,N.es_fingerprint,N.tx_ac,E.tx_acfrom uta_20140210.tx_exon_set_summary_mv Njoin uta_20140210.tx_exon_set_summary_mv E on N.es_fingerprint=E.es_fingerprint and N.tx_ac ~ '^NM_' and E.tx_ac ~ '^ENST' and N.alt_aln_method='transcript' and E.alt_aln_method='transcript';
NCBI (Splign) v. UCSC (BLAT) Alignment StatisticsNCBI (Splign) v. UCSC (BLAT) Alignment StatisticsSplign and BLAT provide significantly different exon structures for 886 transcriptsSplign and BLAT provide significantly different exon structures for 886 transcripts
Are Splignand BLATsimilar ?
31472 (97.3%)transcripts
Y
N
32358transcripts
w/exon structures
➌
886 (2.7%)transcripts
“similar” means either1) identical exon coordinates, or2) coordinates that differ only by short 3' terminal artifacts
19 / 28
Characterization of transcripts discrepanciesCharacterization of transcripts discrepanciesWhether alignments provided by NCBI and UCSC agree with GRCh37 primary sequence.Whether alignments provided by NCBI and UCSC agree with GRCh37 primary sequence.
Splign
BLA
TT F
T 14 18
F 545 311
886 transcripts withsignificant discrepancies
20 / 28
Characterization of transcripts discrepanciesCharacterization of transcripts discrepanciesReference agreement (blue) and alignment “simplicity” (green)Reference agreement (blue) and alignment “simplicity” (green)
Splign
BLA
TT F
T 14 18
F 545 311Splign
BLA
T
T F
T 200(0)
4(97)
F 90(82)
16(84)
Splign
BLA
T
T F
T 6(41)
12(180)
F
Splign
BLA
T
T F
T 434(7)
F 110(652)
Splign
BLA
T
T F
T 14(11)
F
886 transcripts withsignificant discrepancies
21 / 28
ACMG “Must Report” GenesACMG “Must Report” Genes
Green, R. C., Berg, J. S., Grody, W. W., Kalia, S. S., Korf, B. R., Martin, C. L., … Biesecker, L. G. (2013). ACMG recommendations for reporting of incidental findings in clinical exome and genome sequencing. Genetics in Medicine : Official Journal of the American College of Medical Genetics, 15(7), 565–74. doi:10.1038/gim.2013.73
22 / 28
Summary of Splign-BLAT gene-wise coordinate deltas.Summary of Splign-BLAT gene-wise coordinate deltas.
delta # genes # ACMG must report
=0 15206 45
>=1 183 8
>=10 116 0
>=25 6 0
>=50 5 0
>=250 13 0
>=1000 94 3
delta ≝ minimum per gene of maximum per transcript of difference of exon coordinates between NCBI and UCSC.
Example: Variant liftover between transcriptsExample: Variant liftover between transcriptsMapfrom NM_182763.2:c.688+403C>T➀to NC_000001.10:g.150550916G>A➁to ➂ NM_001197320.1:281C>Twith Splign alignments
NM_001197320.1NP_001184249.1
NM_182763.2NP_877495.1
➀
➂
➁
NC_000001.10
26 / 28
Developer InfoDeveloper Info
Testing➢ 91% code coverage➢ 25665 tests variants
● ~200 hand curated, rest from dbSNP
● 23436 sub, 1254 del, 908 ins, 45 delins, 22 dup
● 44 distinct transcripts, many selected for difficulty
➢ >99% concordance with Mutalyzer
● using >100K variants from ClinVar
Upcoming directions(all issues are publicly readable)➢ multi-variant alleles➢ release LRG➢ GRCh38➢ API changes
27 / 28
ConclusionsConclusions
➢ The fidelity of reference-transcript mapping matters● For ~800 transcripts, splign and BLAT generate significantly different
alignments● These differences might affect the interpretation of clinically-relevant
genes (including 3 ACMG must report genes)
➢ Current resources have important limitations
➢ Two tools may help you deal with these limitations● UTA – Freely available archive of transcripts from multiple sources● HGVS – Comprehensive parsing, formatting, manipulation, and validation
of variants
28 / 28
AcknowledgementsAcknowledgements
➢ Invitae● Vince Fusaro● John Garcia● Emily Hare● Kevin Jacobs● Geoff Nilsen● Rudy Rico● Jody Westbrook●
●
● http://goo.gl/dq2uoW
http://bitbucket.com/hgvs/hgvshttp://bitbucket.com/uta/uta➢ Code (Python)➢ Documentation & Examples➢ Issues➢ BED files➢ Code testing is public