Riba:
- Ikon fahimtar RNA-seq gudanawar aiki, nazarin maganganu daban-daban da gyaran gwaji da yawa (FDR) kuma suna da hankali na wucin gadi suna samar da ingantaccen lambar bincike.
- Ikon fassara sakamakon haɓakar hanya a ƙididdiga da ilimin halitta
- Ƙarfin yin amfani da horo na duba zato na ƙididdiga da magudanun gwaji da yawa da kuma tabbatar da mahimmancin ilimin halitta da kansa.
"Omics" shine sunan gamayya don hanyoyin da ke auna duk kwayoyin halitta na tantanin halitta ko nama tare: genomics (duk DNA), transcriptomics (duk RNA / magana), proteomics (duk sunadaran), metabolomics (duk metabolites). Wannan bayanan yana da girma - gwajin RNA-seq ya ƙunshi ma'auni na dubun dubatar kwayoyin halitta. A cikin wannan rukunin, zaku koyi yadda ake amfani da hankali na wucin gadi (AI) a matsayin mataimaki a cikin binciken omics wanda ke rubuta lamba, zaɓi ƙididdiga, da tsara fassarar ilimin halitta; amma za ku koyi dalilin da ya sa dole ne ku tabbatar da sakamakon ƙididdiga da nazarin halittu.
Gargadi mai mahimmanci: A cikin Omics, kurakurai mafi haɗari sune ƙididdiga da ganuwa. AI na iya rubuta lambar da ke ƙetare gyare-gyaren kwatancen da yawa (a ƙasa), zaɓi gwajin da ba daidai ba, ko samar da jerin jerin abubuwan "ƙarya mai inganci". Bugu da ƙari, AI na iya yin da'awar nazarin halittu kamar "wannan kwayar halitta tana ƙaruwa a cikin wannan cuta" ba tare da wata tushe ba. Hanyar da ta dace: don yin bincike tare da aiwatarwa, lambar tantancewa da tabbatar da kowane da'awar ilimin halitta a cikin wallafe-wallafen.
Ka'idoji na asali
- Magana: Nawa ne aka fassara kwayar halitta zuwa RNA; ma'aunin "aikin".
- Bambance-bambancen magana (DE): Halittu waɗanda maganganunsu ke canzawa sosai tsakanin ƙungiyoyi biyu (misali, haƙuri / lafiya).
- p-darajar: Yiwuwar cewa bambanci shine daidaituwa; idan yana ƙarami, ana ɗaukar bambancin "mahimmanci".
- Gyaran kwatance da yawa: Lokacin da aka kalli dubun dubatar kwayoyin halitta a lokaci guda, kwatsam za a sami wasu masu “mahimmanci”. Don gyara wannan, ana amfani da hanyoyin kamar FDR (ƙididdigar binciken karya) / Benjamini-Hochberg. Idan aka bar wannan gyara, ɗaruruwan binciken karya za su tashi.
- log2 ninka canji (log2FC): Logarithm na yanayin magana tsakanin ƙungiyoyi biyu zuwa tushe 2; +1 yana nufin haɓaka ninki biyu, -1 yana nufin raguwa sau biyu.
- Haɓaka Tafarki: Gano waɗanne hanyoyin nazarin halittu (misali rarraba tantanin halitta, rigakafi) ƙwayoyin halittar da aka canza sun tattara; Ana amfani da bayanan bayanai kamar GO da KEGG.
- Tasirin tsari: bambance-bambancen da ba na halitta ba wanda ya taso daga sarrafa samfuran a cikin kwanaki/na'urori daban-daban.
Mataki-mataki: Binciken omics mai ƙarfin AI
1. Bayyana ƙirar gwaji da tambaya. Samfura nawa, rukuni nawa, maimaitawa nawa? Shin ikon ƙididdiga ya isa? Bayyana zane ga AI.
2. Zaɓi kayan aiki / hanya mai dacewa tare da AI. Don RNA-seq, zaɓi daidaitattun hanyoyin kamar DESeq2/edgeR (fakitin ƙididdiga waɗanda aka tsara don ƙidayar bayanai); Yi amfani da ingantattun hanyoyin maimakon kididdigar da AI ta “yi”.
3. Guda lambar kuma duba matsakaicin fitarwa. Kar a ci gaba zuwa bincike na DE ba tare da daidaitawa ba, sarrafa tasirin tsari, maƙasudin inganci (PCA).
4. Ƙaddamar da gyare-gyaren kwatancen da yawa. Tace sakamakon ta hanyar ingantaccen ƙima (padj/FDR), ba ɗanyen p-darajar ba.
5. Tabbatar da fassarar nazarin halittu a cikin adabi. Fassara fitowar wadatar hanyar hanya tare da AI, amma tabbatar da kowane da'awar a tushe.
Tukwici: A cikin bincike na DE, duba inganci da jimillar tasirin jadawali da farko. Idan samfurori sun taru a ranar da aka sarrafa su maimakon ƙungiyar nazarin halittu, yawancin "mahimmancin" kwayoyin halitta da kuke samu suna da tasiri, ba ainihin ilmin halitta ba.
uku mini lokuta
Case 1 - p-darajar da ba a gyara ba. Wani dalibi ya gwada kwayoyin halitta 20,000 tare da lambar da AI ta rubuta kuma ya gano "mahimman kwayoyin halitta 540." Lokacin da ya bincika lambar, ya ga cewa AI ta tsallake gyaran kwatancen da yawa. Lokacin da aka ƙara gyaran FDR, adadin mahimman kwayoyin halitta ya ragu zuwa 32. Ba tare da gyara ba, fiye da 500 na karya za su dogara ne akan labarin.
Hali na 2 - Ƙirƙirar da'awar nazarin halittu. Ga kwayar halitta a cikin jerin DE, wani mai bincike ya tambayi AI "menene wannan kwayar halitta ke yi a cikin wannan cuta?" sai ya tambaya; AI ya bayyana wani tsari mai gamsarwa da labarin. Lokacin da ya bincika kan PubMed, ya ga cewa babu wannan hanyar ko labarin. An cire da'awar daga rahoton.
Case 3 - Tabbatar da samu. Wani ɗalibin PhD ya lura cewa an raba samfuran ta kwana biyu a cikin shirin PCA. An tambayi AI don ƙara ma'anar tasirin batch zuwa lambar; Bayan gyara, jerin kwayoyin halittar sun canza gaba daya kuma sun zama mahimmanci a ilimin halitta. Idan ba tare da ginshiƙi mai sarrafa inganci ba, da an buga sakamakon karya.
Misali: tacewa tare da gyara p-darajar
shigo da pandas azaman pd# bari tebur 'de' ya zama sakamako daga kayan aikin DE (DESeq2/edgeR): # ginshiƙai: gene, log2FC, pvalue, padj (FDR-corrected) de = pd.read_csv ("de_results.csv") mai mahimmanci = de [(de["padj"] & 0 FC = de [(de["padj")] & 0 .[0. 1)] bugawa ("Raw p <0.05:", (de["pvalue") <0.05).sum()) bugawa ("FDR-gyara padj <0.05 & | log2FC|>=1:", len (mahimmanci))
Bambanci tsakanin danyen da lambar da aka gyara yana kwatanta dalilin da yasa kwatancen da yawa ke da mahimmanci.
Samfura huɗu masu kwafi
1) Tsarin nazari da zaɓin hanyar:
Matsayinku: mataimakin bioinformatics. Ƙirƙiri tsarin bincike don gwajin RNA-seq mai zuwa: [yawan ƙungiyoyi, adadin samfura/mai kwafi, tambaya]. Wanne daidaitaccen kayan aiki (DESeq2/edgeR) zan zaɓa kuma me yasa, menene matakan kula da ingancin (PCA, tasirin batch) ake buƙata, ta yaya zan yi amfani da gyaran kwatancen da yawa? Rubuta mataki-mataki.
2) Code + tilas cak:
Rubuta lambar da za a iya aiwatarwa wanda ke aiwatar da bincike na DE mai zuwa: [cikakken bayani]. Dole ne lambar ta haɗa da daidaitawa, ƙirar ingancin PCA, sarrafa tasirin tsari, da gyara FDR (Benjamini-Hochberg). Yi sharhi kowane mataki. Tace tare da gyara p-darajar, ba danyen p-darajar ba.
3) Sharhin inganta hanyar hanya:
Taimaka fassara sakamakon wadatar hanya don wannan jerin mahimman kwayoyin halitta: [jeri/fitarwa]. Bayyana waɗanne hanyoyi ne suka yi fice, amma gaya mani ta wane tushe (GO, KEGG, labarin da aka bita) don tabbatar da kowane da'awar ilimin halitta. Injiniyanci/Lalabaran GWAMNATIN.
4) Binciken kididdiga:
Check the following analysis code for statistical errors: [code]. Specifically: incorrect test selection, omission of multiple comparison correction, ignoring batch effect, insufficient replication. Jera kowace matsala da kuka samu da gyara ta.
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rarrauna: "Nemo mahimman kwayoyin halitta a cikin wannan bayanan RNA-seq."
Matsala: Hanyar, kula da inganci, da kwatancen da yawa ba a ƙayyade ba; AI na iya ba da jerin da ke cike da ƙima ba tare da gyara ba.
Ƙarfafa: "Rubuta lambar da ke yin nazarin DE don wannan ƙididdigar RNA-seq tare da ma'anar DESeq2, ya haɗa da kula da inganci da sarrafa tasiri mai yawa tare da PCA, da tacewa tare da gyaran FDR; sharhi kowane mataki kuma bayyana dalilin da yasa kuka zaɓi wannan hanyar."
Me ya sa yake da iko: Hanyar ita ce ma'auni, inganci da gyara wajibi ne, sakamakon yana iya dubawa.
Hadarin
alama
yin taka tsantsan
tabbataccen ƙarya
Yawancin kwayoyin halitta "ma'ana".
FDR/gyaran kwatancen da yawa
hadin kai tasiri
Ana tattara samfurori da rana
PCA + batch m
gwajin kuskure
Gwaji na yau da kullun don ƙidaya bayanai
Hanyar da ta dace kamar DESeq2/edgeR
halitta ilmin halitta
Injin mara walƙiya
Tabbatar da adabi
rashin isasshen iko
1-2 maimaitawa
Isasshen maimaitawa a cikin ƙira
Kuskuren gama gari
- Tsallake gyaran kwatancen da yawa. Kuskuren ƙididdiga mafi na kowa kuma mafi cutarwa.
- Yin watsi da tasirin gama kai. Yana haifar da bambancin ilimin halitta na ƙarya.
- Aiwatar da gwajin da ba daidai ba don ƙidaya bayanai. RNA-seq yana buƙatar hanyoyi na musamman.
- Karɓar da'awar ilimin halitta ba tare da tushe ba. AI na iya ƙirƙirar kayan aiki da labarai.
- Gabaɗaya tare da ƙarancin maimaitawa. Ba tare da ikon ƙididdiga ba, sakamakon ba shi da tabbas.
Tsanaki: "Mahimmancin ƙididdiga" baya ɗaya da "mahimmancin ilimin halitta." Canji mai ƙanƙanta amma ta fasaha mai mahimmancin canji na iya zama maras muhimmanci a ilimin halitta; A cikin manyan samfurori duk abin da zai iya zama "mahimmanci". Ƙimar log2FC da p-darajar tare.
Zurfi: bangon baya da magudanan kirga sau biyu a cikin wadatar hanya
Sakamakon wadatar hanya ya dogara da zato biyu ɓoye waɗanda yawancin mutane ba su sani ba, kuma AI na iya ƙetare su cikin shiru. Na farko shine zaɓi na baya/universe: haɓakawa yana kwatanta saitin "genes waɗanda suka canza" tare da saitin "waɗanne kwayoyin halitta aka duba". Idan an ɗauki bangon daga gabaɗayan kwayoyin halitta amma gwajin ku kawai yana auna takamaiman kwamiti na nama, sakamakon yana bayyana “an wadata.” Madaidaicin baya sune kwayoyin halitta waɗanda za'a iya bayyanawa/aunawa a zahiri a cikin gwajin. Ƙungiya ɗaya ta sami "mahimman wadatar haɓakar hanyar rigakafi" ta hanyar yin kuskuren ɓoye dukkanin kwayoyin halitta; Lokacin da aka maimaita bincike tare da madaidaicin asali (ma'auni na kwayoyin halitta), wadatar ta ɓace - binciken shine hanyar fasaha.
Na biyu, girman saitin kwayoyin halitta da kirgawa biyu: manya-manya da kuma hanyoyin gaba daya (misali "tsarin tafiyar da rayuwa", dubban kwayoyin halitta) sun bayyana "mahimmanci" a kusan kowane jeri; ƙananan hanyoyi, ƙayyadaddun hanyoyi sun fi sani. Bugu da ƙari, saboda ana samun nau'in kwayar halitta iri ɗaya a hanyoyi da yawa, yana da kuskure don ɗaukar hanyoyin da ke kan gaba a matsayin shaida mai zaman kanta. Batu na uku: baya nuna alkiblar wadata; Ana iya wadatar hanya, amma rabin kwayoyin halittar da ke cikinta na iya karuwa kuma sauran rabi na iya raguwa. Don ganin wannan, ya zama dole a bincika bayanan jagora daban (kamar GSEA).
tarko
alama
yin taka tsantsan
ba daidai ba bango
Komai da alama ya wadatar
Samo bayanan auna kwayoyin halitta
Hanyar gabaɗaya sosai
"Metabolism" kullum yana zuwa
Mayar da hankali kan ƙanana, takamaiman hanyoyi
kirga biyu
hanyoyi masu daidaitawa
Kar a dauke shi a matsayin shaida mai zaman kanta
tsallake hanya
gauraye hawa/sakowa
Gudanar da jagora tare da GSEA
A takaice
- AI a cikin nazarin omics; mataimaki ne wanda ke zaɓar hanyoyin, rubuta lamba, da zayyana sharhi; statistics da nazarin halittu yanke shawara dole ne a barata.
- Daidaitaccen, hanyoyin da aka tabbatar (DESeq2 / EdgeR) yakamata a yi amfani da su; Bai kamata a tsallake kulawar inganci da tantance tasirin haɗin kai ba.
- Gyaran kwatance da yawa (FDR) wajibi ne; ana tace sakamakon tare da ingantaccen darajar.
- Kowane da'awar halitta dole ne a tabbatar da shi a cikin adabi; "mahimmanci" ya kamata a bambanta daga "mahimmanci".
Aikin aikace-aikace
Ɗauki samfurin sakamako DE (ko buɗaɗɗen bayanan RNA-seq). Kwatanta mahimman ƙididdiga masu mahimmanci dangane da ɗanyen p-darajar da gyara ƙofofin padj tare da snippet na sama. Yi sharhi akan bambancin jumla ɗaya. Sannan nemi fassarar ilmin halitta tare da samfuri 3 don fiyayyen halitta kuma duba da'awar da kanku a cikin PubMed.
jerin abubuwan dubawa
- [] Na kimanta ƙirar gwaji da ikon ƙididdiga.
- [ ] Na zaɓi daidaitaccen hanya, dacewa.
- [ ] Na yi PCA da sarrafa ingancin tsari.
- [ ] Na yi amfani da gyare-gyare da yawa (FDR).
- [ ] I filtered the results with corrected p-value.
- [ ] I confirmed the biological claims in the literature.