De novo iden)fica)on of SNPs from RNA- seq data in non- model species

Transcription

1 De novo iden)fica)on of SNPs from RNA- seq data in non- model species Hélène Lopez- Maestre 8th Novembre 2016

2 Why work with RNAseq? Lower cost SNPs from expressed regions SNPs with a more direct func:onal impact

3 How to detect SNPs in RNAseq data? for non model species?

4 reads from RNAseq SNP iden:fica:on from RNAseq with a reference genome

5 reads from RNAseq SNP iden:fica:on from RNAseq 1) with a reference genome reference genome

6 reads from RNAseq SNP iden:fica:on from RNAseq 1) with a reference genome reference genome Several tools : SNPiR GATK SAMtools mpileup etc.

7 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : Trinity Oases

8 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

9 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

10 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

11 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

12 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

13 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

14 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

15 reads from RNAseq SNP iden:fica:on from RNAseq 2) with an assembled transcriptome transcriptome assembly : { two assembled transcripts from one gene due to alterna)ve splicing A A C B D D

16 SNP iden:fica:on from RNAseq 3) from a De Bruijn graph reads from RNAseq local assembly

17 SNP iden:fica:on from RNAseq 3) from a De Bruijn graph ATTCCTGCAATAC ATTCCTACAATAC GGTAGGTACATCGTGT GTTGTGCATGAATTCTG TGCATGAATTCTACAATA ATGAGGTACATGCAT GCATGAATTCTGCAATAC GTTGTGCATGAATTCTA KisSplice DiscoSNP Bubble of size = 2k+1

18 Kissplice Filtres Sequencing errors Inexacts repeats

19 Sequencing errors low expression 39G 1A high expression 390G 10A ) absolute cutoff rela)ve cutoff

20 Inexacts repeats...attgctgcaattc...attgctacaattc......attgctgcaattc...attgctacaattc...

21 Inexacts repeats...attgctgcaattc...attgctacaattc......attgctgcaattc...attgctacaattc......attgctgcacttc...attgctgcaattc...attgctacaattc......attgctgcacttc...attgctgcaattc...attgctacaattc...

22 Inexacts repeats...attgctgcaattc...attgctacaattc......attgctgcaattc...attgctacaattc......attgctgcacttc...attgctgcaattc...attgctacaattc......attgctgcacttc...attgctgcaattc...attgctacaattc... X10 copies

23 SNP iden:fica:on from RNAseq 1) with a reference genome 2) with an assembled transcriptome 3) directly from the De Bruijn graph

24 Valida:on of the method on real data Human dataset genome and annota)on available 1000G project : genotype of 1000 individuals Geuvadis : RNAseq of these individuals RNAseq from 2 x 5 central europeans and 2 x 5 toscans

25 Valida:on of the method on real data Recall : Out of all the «True» SNPs, the propor)on predicted by the method Precision : Out of all the predicted SNPs, the propor)on that is correctly predicted

26 True SNPs All the SNPs according to the genotype of the 20 indivuals (1000G project) locus covered by a minumum of 5 reads

27 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C= True SNPs minimum Coverage

28 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

29 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

30 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

31 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

32 Recall and precision Recall GATK genome MP transcriptome MPlong transcript. Kp default Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

33 Recall and Allele Frequency MAF : Minor Allele Frequency All MAF MAF > 0.05 MAF > 0.15 Recall GATK genome MP transcriptome MPlong transcript. MPcdhit transcript. Kp default Kp C=0.02 Recall Recall True SNPs minimum Coverage True SNPs minimum Coverage True SNPs minimum Coverage

34 Downstream analyses

35 Downstream analyses i) Variant impact on the amino- acid? ii) Variant specific to a biological condi)on?

36 Predic:on of the amino acid changes reads from RNAseq transcriptome assembly

37 Predic:on of the amino acid changes KisSplice2refTranscriptome transcriptome assembly

38 Predic:on of the amino acid changes KisSplice2refTranscriptome transcriptome assembly

39 Predic:on of the amino acid changes KisSplice2refTranscriptome transcriptome assembly

40 Predic:on of the amino acid changes KisSplice2refTranscriptome C G A T transcriptome assembly

41 Predic:on of the amino acid changes KisSplice2refTranscriptome Arginin C G A T His)din transcriptome assembly

42 Quan:fica:on

43 Quan:fica:on 25 22

44 Condi:on specific variant Condi)on 1 Condi)on 2 Condi)on 1 23 Condi)on Condi)on 1 18 Condi)on 2

45 Condi:on specific variant Condi)on 1 Condi)on 2 Condi:on 1 23 Condi)on 2 2 KissDE 4 Condi)on 1 18 Condi:on 2

46 Pipeline : Input fastq files D00393:39:C3PHDACXX:2:1101:8716:2263 1:N:0:CTATAC TCAGTCTTTTTCTCTCTCCTAGTCCCATAATGCCTTGACGAACATCAGTCTGTCCCATAATGCTCTCCTATTA + D00393:39:C3PHDACXX:2:1101:11867:2372 1:N:0:CTATAC TACGTCGGAAATATATAAAAAAATGGCACTGGAAGAATTCACTGGCACTGGAAGAATTCATGGCACTGGA + D00393:39:C3PHDACXX:2:1101:3047:2629 1:N:0:CTATAA CCGTGGTCCCATAATGCCTTGACACGGTCTTGGCACTGGAAGAATTCAAGAATTCATGGCACTGGAAGAA + D00393:39:C3PHDACXX:2:1101:8009:2719 1:N:0:CTATAC CCCGACTGAAGGTGTACTCATGGCCAGGTCCCATAATGCCTTGACCTCTCCTATTACTGAAGGGGTGTACT + CCCCCCCCCCCCCCCCCBCCCCCCCCCCBC?CBBBBBBBBBBBBBCFFFFFGGHHFHHHHHHHHJJIIJIIJJJI

47 Pipeline : Output KisSplice + KissDE + Kissplice2refTranscriptome output : c54770_g1_i bcc_487 Cycle_259 CAA CCA Q P NonSyn. No

48 Pipeline : Output KisSplice + KissDE + Kissplice2refTranscriptome output : c54770_g1_i bcc_487 Cycle_259 CAA CCA Q P NonSyn. No c54495_g1_i bcc_6275 Cycle_1 GCA GCC A A Syn. No

49 Pipeline : Output KisSplice + KissDE + Kissplice2refTranscriptome output : c54770_g1_i bcc_487 Cycle_259 CAA CCA Q P NonSyn. No c54495_g1_i bcc_6275 Cycle_1 GCA GCC A A Syn. No c59969_g1_i bcc_9888 Cycle_0 AGC TGC S C NonSyn. Yes

50 Pipeline : Output KisSplice + KissDE + Kissplice2refTranscriptome output : c54770_g1_i bcc_487 Cycle_259 CAA CCA Q P NonSyn. No c54495_g1_i bcc_6275 Cycle_1 GCA GCC A A Syn. No c59969_g1_i bcc_9888 Cycle_0 AGC TGC S C NonSyn. Yes c65293_g5_i2 847 bcc_6446 Cycle_0 N/A N/A N/A N/A NonCod. Yes

51 Experimental Valida:ons

52 Asobara tabida ovaries infected with Wolbachia

53 Asobara tabida ovaries infected with Wolbachia 2 lines with variable dependency

54 Asobara tabida ovaries infected with Wolbachia 2 lines with variable dependency RNAseq (ovaries) pool of 30 individuals

55 Asobara tabida ovaries infected with Wolbachia 2 lines with variable dependency RNAseq (ovaries) pool of 30 individuals SNPs of which condi)on specific 36/36 SNPs validated (rt- PCR + sequencing) of which 7 predicted by Kissplice only

56 Take- home message Detec)ng SNPs in non model species from RNAseq Similar performances (precision and recall) compared to methods using a ref. genome Precision for SNP discovery : >80% Test for condi)on specific variant Precision : 96% Experimental valida)ons Lopez- Maestre et al. NAR h^p://kissplice.prabi.fr/twas/

57 Perspec:ves Improve the detec)on of close SNPs and SNPs in highly polymorphic regions : increase KisSplice recall «True» SNPs hard to detect for all methods

58 Acknowledgments ARN Colibread L. Brinza C. Marchet J. Kielbassa S. Bas)en M. Bou)gny D. Monin C. Vieira F. Picard N. Kremer F. Vavre M. Sagot V. Lacroix G. Sacomoto

59 Experimental valida:on * * * SFR2 SFR3 Nitric oxide synthase SFR2: 0% / SFR3: 100% Protein ovarian tumor SFR2: 0% / SFR3: 33% Pyrimidodiazepine synthase SFR2: 100% / SFR3: 67%

60 Mul:ple SNPs Exclus de l étude : Fort taux de faux posi)fs Probablement des répé))ons inexactes

61 Close SNPs : phased distance <k Bubble of size >2k+1

62 Close SNPs : non- phased

63 Recall vs MAF All MAF MAF > 0.05 MAF > 0.15 Recall GATK genome MP transcriptome MPlong transcript. MPcdhit transcript. Kp default Kp C=0.02 Recall Recall True SNPs minimum Coverage True SNPs minimum Coverage True SNPs minimum Coverage

64 Kissplice parameters Recall MPlong Kp default Kp k=31 Kp k=51 Kp b=10 Kp C=0.10 Kp C=0.02 Precision True SNPs minimum Coverage Predicted SNPs minimum coverage

65 Impact on amino acids

66 Impact on amino acids

67 Impact on amino acids Precision of impact on aa : 96%

68 Impact on amino acids Precision of impact on aa : 96% Precision of the pipeline > 77%

69 KissDE C1 C2 Var Var % 0%

70 KissDE C1 C2 C1 C2 Var Var Var Var % 0% 100% 30%

71 KissDE C1 C2 C1 C2 Var Var Var Var % 0% 100% 30% C1 C2 Var Var % 30%

72 Difference in allele frquency

73 Difference in allele frquency

74 Difference in allele frquency Pvalue 0.05 #reads 100

75 KissDE model GLM : log(λ ijk )= μ + α i + β j log(λ ijk )= μ + α i + β j + (αβ) ij