理解して使うRNA Velocity解析ツール-最近のツール編

2023/02/08 Bio"Pack"athon 2023 #2 発表資料

Transcript

1. ཧղͯ͠࢖͏RNA Velocityղੳπʔϧ - ࠷ۙͷπʔϧฤ - Mika Yoshimura 2023/02/08 ཧԽֶݚڀॴ
2. scVelo։ൃνʔϜʹΑΔϨϏϡʔ࿦จ • RNA velocity—ݱࡏͷ՝୊ͱকདྷͷల๬ Volker Bergen, Ruslan A Soldatov, Peter V Kharchenko, Fabian J Theis Mol Syst Biol. 2021 Aug;17(8):e10282 • Velocityղੳख๏ͷݶքʹ͍ͭͯͷ࣮ྫͱFeature PlansΛࣔ͢
3. scVeloͷࠓޙͷల๬ A. velocityϞσϧͷ֦ு ࡉ๔ͷঢ়ଶؒ΍μΠφϛοΫͳϓϩηεதʹ͓͜Δ ύϥϝʔλͷಈతͳมԽΛߟྀ֦ͨ͠ு B. ࡉ๔ಛҟతμΠφϛΫεͷ֬཰࿦తϞσϧ RNA velocityͷϞσϧ͸ܾఆ࿦తͳඍ෼ํఔࣜͰ هड़͞ΕΔͨΊɺෆۉҰੑ͸ແࢹ͞Ε͍ͯΔ → ֬཰తมಈΛϞσϧʹऔΓࠐΉ C. γεςϜμΠφϛΫεʹ޲͚ͨଟมྔղੳϞσϧ RNA velocityͷϞσϧͰ֤Ҩ఻ࢠͷ੍ޚؔ܎͸ແࢹ͞ΕΔ → ࡉ๔ঢ়ଶͷભҠʹԊ੍ͬͨޚੑ૬ޓ࡞༻Λهड़͢Δ ଟมྔϞσϧʹ֦ு͢Δ D. Multi-modal omics models protein, ࡉ๔࣭ͷmRNA, ΩϟοϐϯάmRNA, ΫϩϚνϯͷঢ়ଶ౳ɺଞͷγάφϧΛϞσϧʹ૊ΈࠐΉ
4. scVeloͷݱࡏ ෳ਺ͷܥ౷Λ΋ͭࡉ๔ूஂ ಛఆͷΫϥελͰɺਪఆ͞ΕͨμΠφϛΫε(ࢵͷۂઢ) Ͱ͸આ໌Ͱ͖ͳ͍ڍಈΛࣔ͢Ҩ఻ࢠ ҟͳΔڍಈͷΫϥελ Velocity࠶ܭࢉ
5. [ٙ໰] 2022.12 ࣌఺Ͱɺ Velocityղੳͷ՝୊Λೝࣝͨ͠ ৽͍͠Velocityπʔϧ͸ ͲΕ͘Β͍ൃද͞Ε͍ͯΔͷ͔ʁ → ϨϏϡʔ࿦จ Mol Syst Biol. 2021 ΛҾ༻͢Δ࿦จΛௐࠪ
6. V.Bergen.et.al, Mol Syst Biol. 2021 ΛҾ༻͍ͯ͠Δ࿦จ 22೥12݄࣌఺ Ҿ༻࿦จ਺ : 54ใ ॏෳΛআڈ : 51ใ 23೥2݄ݱࡏ → 70ใ A : Review(ղੳऀ޲͚) = 1ຊ B : ख๏ݕূ= 2ຊ C : طଘvelocityͷԠ༻ = 3ຊ D : scVeloͷޙܧख๏ = 13ຊ E : scVeloͷFigΛܝࡌ = 10ຊ F : velocityʹݴٴɾଞ = 21ຊ
7. A : Review(ղੳऀ޲͚) = 1ຊ B : ख๏ݕূ= 2ຊ C : طଘvelocityͷԠ༻ = 3ຊ D : scVeloͷޙܧख๏ = 13ຊ E : scVeloͷFigΛܝࡌ = 10ຊ F : velocityʹݴٴɾଞ = 21ຊ V.Bergen.et.al, Mol Syst Biol. 2021 ΛҾ༻͍ͯ͠Δ࿦จ 22೥12݄࣌఺ Ҿ༻࿦จ਺ : 54ใ ॏෳΛআڈ : 51ใ 23೥2݄ݱࡏ → 70ใ
8. scTour DeepVelo LatentVelo PhyloVelo Pyro-Velocity veloVI VeloVAE spliceJAC Molecular Systems Biology (2022)18:e111 MultiVelo UniTVelo κ-velo/eco-velo ValoAE Nature Biotechnology (2022) Nature Comm volume 13, 6586 (2022) Cell volume 185, Issue 4, 2022 PLoS Comput Biol. 2022 Sep 28 Proc Natl Acad Sci U S A. 2021 Dec 7 Preprints Published not use “velocity” scVeloͷޙܧख๏ Ұཡ
9. Ϟσϧ֦ு MultiVelo Nature Biotechnology (2022) UniTVelo Nature Comm volume 13, 6586 (2022) Ϟσϧ֦ு + embeddingͷվྑ κ-velo/eco-velo Cell volume 185, Issue 4, 2022 PLoS Comput Biol. 2022 Sep 28 ม෼ΦʔτΤϯίʔμʔ(ੜ੒Ϟσϧ) ValoAE Proc Natl Acad Sci U S A. 2021 Dec 7 scTour LatentVelo Pyro-Velocity ਂ૚ੜ੒Ϟσϧ veloVI άϥϑ৞ΈࠐΈ ωοτϫʔΫ DeepVelo ม෼ϕΠζ VeloVAE ࡉ๔ܥ౷थ ଟมྔεϓϥΠγϯάϞσϧ (not use velocity) PhyloVelo spliceJAC Molecular Systems Biology (2022) ΦʔτΤϯίʔμʔ
10. Ϟσϧ֦ு MultiVelo Nature Biotechnology (2022) UniTVelo Nature Comm volume 13, 6586 (2022)
11. MultiVelo Nature Biotechnology (2022) ɾΤϐήϊϜσʔλΛऔΓೖΕͨඍ෼ํఔࣜϞσϧ ɾసࣸ଎౓͸ΫϩϚνϯΞΫηγϏϦςΟ (c)ʹ ൺྫ࣌ؒ͠తʹมԽ͢ΔͱԾఆ ɾ࣌ؒt͕ਐΉʹͭΕͯมԽ͢Δঢ়ଶkΛϞσϧԽ ɾҨ఻ࢠൃݱͷεΠον࣌ؒ (ΫϩϚνϯด؀ɺసࣸ༠ಋɺసࣸ཈੍͕࢝·ΔλΠϛϯά) ɾҨ఻ࢠݻ༗ͷVelocityύϥϝʔλ → 3࣍ݩҐ૬ϙʔτϨʔτΛ༻͍ͯਪఆ
12. UniTVelo Nature Comm volume 13, 6586 (2022) ɾspliced-RNAΛ࢖༻͠Ҩ఻ࢠൃݱͷ׬શͳμΠφϛΫεΛ ์ࣹঢ়جఈؔ਺ (RBF) ͰϞσϧԽɺτοϓμ΢ϯతʹRNA velocityΛਪఆ͢Δ ɾεϓϥΠγϯά͞Ε͍ͯͳ͍RNA͕શ͘ͳ͍ঢ়ଶͰ΋VelocityΛ༧ଌ͢Δ͜ͱ͕Ͱ͖Δ ɾࡉ๔͝ͱʹτϥϯεΫϦϓτʔϜશମͰڞ༗͢Δજࡏtime pointΛಋೖ͠ɺ Ҩ఻ࢠؒͰͷϊΠζ(ํ޲ੑͷෆҰக)Λղܾ → ࡉ๔෼Խܦ࿏ͷ༧ଌਫ਼౓͕޲্
13. Ϟσϧ֦ு + embeddingͷվྑ κ-velo/eco-velo Cell volume 185, Issue 4, 2022 PLoS Comput Biol. 2022 Sep 28
14. Cell volume 185, Issue 4, 2022 ɾ୅ँඪࣝγʔέϯεσʔλΛ spliced/unsplicedμΠφϛΫεʹجͮ͘Ϟσϧʹ౷߹͠ ਅͷ(absolute) velocity Λਪఆ ɾ୅ँϥϕϦϯάͳ͠ͷ1ࡉ๔RNA-seqσʔλʹ΋ద༻Մ ɾ֤ࡉ๔ͷҨ఻ࢠൃݱྔͱมԽ଎౓(velocity) ͔Βɺ Ҩ఻ࢠൃݱྔϕΫτϧΛड͚औͬͯมԽ଎౓Λฦ͢ ؔ਺ f (ϕΫτϧϑΟʔϧυ)Λֶश͢Δ ɾ࠶ߏ੒͞ΕͨղੳతϕΫτϧϑΟʔϧυΛ༻͍ͯɺ ࡉ๔͕ଘࡏ͠ͳ͍ྖҬʹରͯ͠΋ɺ௿࣍ݩۭؒͷ ϕΫτϧϑΟʔϧυΛ਺ֶతʹ෮ݩͨ͠Γɺ ࡉ๔ӡ໋มભɺҨ఻ࢠ৙ཚͷ݁ՌͳͲͷ༧ଌ͕Մೳ ࢀߟɿ1ࡉ๔γʔέϯγϯάσʔλΛ༻͍ͨҨ఻ࢠൃݱมԽ༧ଌ (in silico perturbation) ख๏ (Dynamo, SCENIC+) ͷ঺հ https://qiita.com/masa12/items/1e564768ad98c0d4cf9c
15. κ-velo/eco-velo PLoS Comput Biol. 2022 Sep 28 ɾϝΠϯ͸κ-velo ɾࡉ๔ີ౓ = Ҩ఻ࢠʹґଘ͠ͳ͍άϩʔόϧͳ જࡏ࣌ؒ “εέʔϦϯάϑΝΫλʔκ”ΛϞσϧʹಋೖ (εέʔϧෆมੑ໰୊Λࠀ෰) ɾඇઢܗ௿࣍ݩۭؒembeddingʹ NyströmࣹӨΛ༻͍Δ ɾeco-velo͸ ߴ࣍ݩͷRNA velocityΛਪ࿦͠ͳ͍ ώϡʔϦεςΟοΫͳΞϓϩʔν
16. ม෼ΦʔτΤϯίʔμʔ(ੜ੒Ϟσϧ) ValoAE Proc Natl Acad Sci U S A. 2021 Dec 7 scTour LatentVelo Pyro-Velocity ਂ૚ੜ੒Ϟσϧ veloVI άϥϑ৞ΈࠐΈ ωοτϫʔΫ DeepVelo ม෼ϕΠζ VeloVAE ΦʔτΤϯίʔμʔ
17. VeloAE Proc Natl Acad Sci U S A. 2021 Dec 7 ɾϊΠζͷଟ͍ߴ࣍ݩspliced/unsplicedϕΫτϧΛ ΦʔτΤϯίʔμͰ௿࣍ݩજࡏۭؒʹࣹӨ͠ɺ ઢܗճؼ(=ఆৗঢ়ଶϞσϧ)Λ༻͍ͯજࡏ্ۭؒͷ velocityΛਪఆ͢Δ ※embedding͸scVeloͱಉ͘͡ભҠ֬཰Λ༻͍Δ scTour ɾม෼ΦʔτΤϯίʔμ(VAE)ͱneural ODEΛ༻͍ͯɺ ڭࢣͳ͠ͰpseudotimeɺτϥϯεΫϦϓτʔϜϕΫτϧ৔ɺ જࡏۭؒΛಉ࣌ʹਪఆ ɾόονޮՌΛड͚ͣɺࣄલ஌ࣝ(ϧʔτࡉ๔ͷࢦఆͳͲ)΍ spliced/unspliced mRNAͷ۠ผ΋ඞཁͳ͍ biorxiv.org/content/10.1101/2022.04.17.488600v2.full
18. LatentVelo ɾม෼ΦʔτΤϯίʔμ(VAE)Ͱࡉ๔ͷΧ΢ϯτσʔλΛ ௿࣍ݩજࡏۭؒʹຒΊࠐΈɺજࡏ্ۭؒͰͷ ෼ԽμΠφϛΫεΛneural ODEͰهड़͢Δ ɾVeloAEͱҟͳΓɺࡉ๔ঢ়ଶΛֶशͨ͠જࡏۭؒʹຒΊࠐΈ μΠφϛΫεΛࡉ๔ঢ়ଶ΍ܥේʹґଘͤ͞Δ͜ͱ͕Ͱ͖Δ ɾ୯ʹہॴతͳVelocityϕΫτϧΛग़ྗ͢ΔͷͰ͸ͳ͘ɺ ݸʑͷࡉ๔͝ͱʹάϩʔόϧͳܦ࿏ Λਪఆ͢Δ → જࡏతͳيಓ΋༧ଌ͢Δ͜ͱ͕Ͱ͖Δ ɾࡉ๔ঢ়ଶͱRNA velocityʹର͢ΔҰׅόονิਖ਼͕Մೳ biorxiv.org/content/10.1101/2022.08.22.504858v2.full
19. Pyro-Velocity ɾspliced/unsplicedϦʔυΧ΢ϯτΛ༻͍ͯvelocityϕΫτϧϑΟʔϧυ & ڞ༗જࡏ࣌ؒͷෆ࣮֬ੑਪఆΛߦ͏֬཰తੜ੒Ϟσϧ ɾม෼ਪ࿦ͷࣗಈԽͱͯ͠ఏҊ͞Ε͍ͯΔࣗಈඍ෼ม෼ਪ࿦(Automatic Differentiation Variational Inference ; ADVI)ͱ͍͏ ΞϧΰϦζϜΛར༻͠ɺۙࣅࣄޙ෼෍Λಋग़͢Δ ɾvariable genesͷબ୒ɺओ੒෼ͷ਺ɺkNNʹΑΔฏ׈ԽͳͲͷલॲཧεςοϓ΍ɺҨ఻ࢠ͝ͱͷಠཱͨ͠velocityʹର͢Δ ฏ׈Խɺू໿ͳͲΞυϗοΫͳޙॲཧ͕ඞཁͳ͍ biorxiv.org/content/10.1101/2022.09.12.507691v2.full Pyro : pythonͷػցֶशϥΠϒϥϦPyTorchΛ ϕʔεʹͨ֬͠཰తϓϩάϥϛϯάݴޠ
20. veloVI ɾscVeloͱಉ͡ϥϘ͔Βͷ࿦จ ɾ ਂ૚ੜ੒Ϟσϧ(Deep Generative Model) Λ༻͍ͯɺ ൓Ԡ଎౓ύϥϝʔλͱજࡏ࣌ؒΛֶश͠ͳ͕Βɺ ͢΂ͯͷࡉ๔ͱҨ఻ࢠؒͰ৘ใΛڞ༗͢ΔϞσϧΛհͯ͠ RNA velocityͷਪ࿦Λ࠶ఆࣜԽ ɾvelocityͷܦݧతࣄޙ෼෍(ࡉ๔-Ҩ఻ࢠϨϕϧ)Λग़ྗ → ಺ࡏతͳෆ࣮֬ੑ + ࡉ๔ͷۙ๣ؒͷมಈੑͷ྆ํΛ ఆྔԽͰ͖Δ https://github.com/YosefLab/velovi ɾ਺ϲ݄Ҏ಺ʹscvi-tools ʹҠߦ + scVeloͰϥούʔ͕Ϗϧυ biorxiv.org/content/10.1101/2022.08.12.503709v1.full
21. DeepVelo ɾάϥϑ৞ΈࠐΈωοτϫʔΫ(GCN)Λ༻͍ͯɺҨ఻ࢠ͝ͱɺࡉ๔͝ͱͷRNAεϓϥΠγϯάɾ෼ղ଎౓Λਪఆ ɾҟछࡉ๔ूஂ΍ϚϧνϦωʔδΛ࣋ͭσʔλʹରͯ͠ɺෳࡶͳ෼ԽಈଶͷRNA velocityΛϞσϧԽ͢Δ͜ͱ͕Մೳ biorxiv.org/content/10.1101/2022.04.03.486877v2.full
22. VeloVAE ɾม෼ϕΠζΛ༻͍ͯɺ֤ࡉ๔ͷજࡏ࣌ؒɺࡉ๔ͷঢ়ଶɺ ൓Ԡ଎౓ύϥϝʔλͷࣄޙ෼෍Λਪఆ ɾσʔλద߹ੑͱ෼Խํ޲ͷਪఆਫ਼౓ͷ఺Ͱɺ ैདྷͷΞϓϩʔνΛେ෯ʹ্ճΔ ɾࡉ๔ঢ়ଶม਺ͱશҨ఻ࢠͰڞ༗͢Δજࡏ࣌ؒͷಋೖʹΑͬͯ ैདྷͷख๏Ͱ͸ଊ͑ΒΕͳ͔ͬͨൃݱμΠφϛΫεͷ ఆੑతಛ௃ΛϞσϧԽͰ͖Δ (஗Ԇ༠ಋɺॳظ཈੍ɺసࣸϒʔετɺ෼ذͳͲ) biorxiv.org/content/10.1101/2022.07.08.499381v2.full
23. ࡉ๔ܥ౷थ ଟมྔεϓϥΠγϯάϞσϧ (not velocity) PhyloVelo spliceJAC Molecular Systems Biology (2022)
24. PhyloVelo ɾҰࡉ๔τϥϯεΫϦϓτʔϜ + ܥ౷थ৘ใ Ͱ ࡉ๔ӡ໋ͷભҠΛܥ౷తʹϚοϐϯά͢Δख๏ ɾࡉ๔෼྾ʹ൐͏୯ௐൃݱҨ఻ࢠʢMEGsʣ͸ɺ ຤୺ࡉ๔ͷछྨʹؔΘΒͣɺܥ౷෼ذʹԊͬͯ ࿈ଓతʹൃݱ͕૿ݮ͢Δ ɾMEGΛར༻͠ɺࡉ๔ܥ౷थʹ෼ղ͞Εͨ scRNA-seqσʔλ͔ΒvelocityϑΟʔϧυΛ ௿࣍ݩ্ۭؒͰޙΖ޲͖ʹ࠶ߏங͢Δ biorxiv.org/content/10.1101/2022.10.06.511094v2.full
25. spliceJAC Molecular Systems Biology (2022) A. unspliced mRNAͱspliced mRNAͷΧ΢ϯτɺ ࡉ๔ͷΞϊςʔγϣϯΛೖྗͱ͢Δ B,C. ଟมྔmRNAεϓϥΠγϯάϞσϧʹΑΓ ֤ࡉ๔ঢ়ଶʹಛҟతͳҨ఻ࢠ-Ҩ఻ࢠ૬ޓ࡞༻ߦྻΛग़ྗ D. ࡉ๔ঢ়ଶؒͷ੾Γସ͑Λɺঢ়ଶಛҟతͳ Ҩ఻ࢠ੍ޚωοτϫʔΫʹΑΔߴ࣍ݩϥϯυεέʔϓ্ͷ ભҠͱͯ͠ղऍ͢Δ E. ҆ఆੑղੳʹΑΓɺࡉ๔ঢ়ଶؒͷҠߦʹॏཁͳ “ભҠҨ఻ࢠ “Λ༧ଌͰ͖Δ (F. ֤ࡉ๔ঢ়ଶ͸ɺ੍ޚωοτϫʔΫʹัଊ͞ΕͨҨ఻ࢠؒͷ ಛҟతͳ૬ޓ࡞༻Λࣔ͢)
26. ·ͱΊ ɾscVelo։ൃνʔϜ͕Mol Syst Biol. 2021ʹͯ໰୊ఏىͨ͠՝୊ʹ͍ͭͯɺޙܧख๏͸ͦͷଟ͘Λ ࠀ෰ͭͭ͋͠Δ ݸʑͷख๏Ͱͷ՝୊ ɾલॲཧ → ਖ਼֬ͳunspliced RNAͷఆྔͳͲ ɾطଘϞσϧͷ֦ு → ֬཰తμΠφϛΫεͷΑΓྑ͍ਪ࿦ ɾੜ੒Ϟσϧͷ׆༻ → ϋΠύʔύϥϝʔλͷௐ੔ ɾ΄ͱΜͲ؍ଌ͞Εͳ͍Ұաੑͷࡉ๔ΫϥελΛ࣋ͭɺૄͳࡉ๔ܥේूஂͰͷVelocityਪఆ etc… ݱঢ়ɺͲͷख๏΋ैདྷͷscVeloͱ͔͠ൺֱ͍ͯ͠ͳ͍ͨΊɺϕϯνϚʔΫ࿦จ͕଴ͨΕΔ scVeloνʔϜ΋ޙܧͷVeloVIΛ։ൃ͓ͯ͠ΓɺࠓޙVelocityղੳͷඪ४͸Ξοϓσʔτ͍ͯ͘͠ͱ༧૝