Riba:
- Ikon fahimtar matsalar kwatancen da yawa kuma ya haɗa da gyaran FDR (ƙididdigar binciken karya) a cikin bincike
- Ability don bambance ƙididdiga da mahimmancin ilimin halitta ta hanyar kimanta p-darajar da girman sakamako (canjin log2 fold) tare
- Ikon ganewa da guje wa tarkon bayanai masu girma kamar tasirin batch da tsalle-tsalle
Kalmar "omics" tana bayyana hanyoyin da ke auna dukkanin nau'ikan kwayoyin halitta a cikin tantanin halitta: genomics (duk DNA), transcriptomics (duk RNA / maganganun kwayoyin halitta), proteomics (duk sunadaran), metabolomics (dukkan kananan kwayoyin halitta). Alamar gama gari na waɗannan ma'auni shine girman girman su: dubunnan ko ma dubun dubatar masu canji (genes, proteins) ana auna su lokaci guda a cikin samfuri ɗaya, amma adadin samfuran yawanci ƙananan (misali marasa lafiya 20). Wannan halin da ake ciki "masu yawa masu yawa, ƙananan samfurori" shine tushen kalubale na musamman ga ilmin halitta da kuma yankunan da AI zai iya taimakawa.
A cikin wannan rukunin, za mu tattauna nazarin bambance-bambancen magana (neman kwayoyin halitta waɗanda maganganunsu ke canzawa sosai tsakanin ƙungiyoyi biyu) da kuma rawar da basirar wucin gadi a cikin wannan aikin ta hanyar misali na transcriptomics (RNA-seq).
Babban matsalar manyan bayanai masu girma
Idan ka gwada dubban kwayoyin halitta a lokaci daya, za ka sami kwayoyin halitta da suke da alama "mahimmanci" kwatsam, ko da babu bambance-bambance na gaske. Idan kun gwada kwayoyin halitta 20,000 tare da gefen kuskure 5%, ~ 1,000 genes na iya zama "mahimmanci" kwatsam. Wannan shi ake kira matsalar kwatantawa da yawa kuma ita ce mafi muni a cikin binciken omics. Magani shine gyara p-darajar (misali lissafin FDR - ƙimar gano ƙarya tare da hanyar Benjamini-Hochberg). AI yana taimakawa sosai wajen bayyana wannan ra'ayi da rubuta lambar da ta dace; amma alhakinku ne ku tuna kuyi amfani da gyaran.
Tukwici: Idan ka ga lamba kamar "kwayoyin halitta 3,000 sun canza sosai" a sakamakon omics, ka firgita. Wannan yawanci alama ce ta cewa ba a yi gyaran kwatance da yawa ba. Jeri na gaske zai kasance dubun zuwa ɗari da yawa kwayoyin halitta a cikin ingantaccen gwaji.
RNA-seq bambancin magana: mataki-mataki
- Raw ƙidaya: Tebu mai ɗauke da adadin karantawa ya faɗi a cikin kowane samfurin ga kowane jinsin halitta.
- Quality da tacewa: Yi watsi da kwayoyin halitta tare da ƙarancin magana.
- Daidaitawa: Daidaitaccen girman ɗakin karatu tsakanin samfurori (lambar ƙima ba ta kwatankwaci).
- Samfurin ƙididdiga: Gwaji bambancin rukuni tare da DESeq2 ko EdgeR (Libraries R) ko pyDESeq2 a cikin Python.
- Gyaran kwatance da yawa: ƙididdige FDR; yawanci madaidaicin FDR <0.05.
- Girman tasiri: Yi ƙima tare da canjin log2 ninka: sau nawa magana ta ƙaru/raguwa.
- Sharhi: Haɗa mahimman kwayoyin halitta tare da hanyoyin nazarin halittu.
Hankali na wucin gadi 3-6. Yana ƙididdige matakan matakai, yana bayyana ra'ayoyi, kuma yana taimaka muku fassara abin fitarwa. Duk da haka, samfurin ba zai iya cewa "wane gene ya canza" ba tare da ganin danyen bayanan ku ba; Lambar da ƙididdiga sun gaya muku wannan.
Samfuran gaggawar da za a iya kwafi
Matsayi: Kai ne mataimaki na bincike na rubutu. Magana: Ina da RNA-seq raw count table (CSV) daga sarrafawa 12, samfuran jiyya 12. Aiki: Jera matakan bincike na bambance-bambancen magana tare da pyDESeq2, bayyana dalilin da yasa kowane mataki ya zama dole. Shirya farko, lamba na biyu. Tabbatar kun haɗa da gyaran kwatancen da yawa.
Sakamakon bincike na ya nuna kwayoyin halitta 4,200 a matsayin "p<0.05". Me yasa hakan zai iya zama abin tuhuma? Bayyana gyare-gyaren kwatance da yawa (Benjamini-Hochberg FDR) kuma ba Python lambar da ke yin tacewa daidai.
Rubuta lambar da za ta zana mahallin dutsen mai aman wuta daga tebur na sakamako na banbanta (gene, log2FC, ginshiƙan padj). Launi kwayoyin halitta tare da FDR <0.05 da |log2FC|>1, yi lakabin saman 10.
Ta yaya zan bincika wannan muhimmin jeri na gado don wadatar hanya? Yi bayanin gseapy ko g:Profiler matakan. Kar a da'awar cikakken dalili a cikin sharhin, yi amfani da yaren daidaitawa. Genelist: [jerin]
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rauni: "Ku gaya mani waɗanne kwayoyin halitta suke da mahimmanci a yawan amfanin RNA-seq."
Karfi: "Ina da pyDESeq2 fitarwa daga 12 iko, 12 magani samfurori: tebur tare da gene, log2FoldChange, padj ginshikan. Ba da code cewa tace muhimman kwayoyin halitta tare da kofa na FDR <0.05 da | log2FC|> 1, bayar da rahoton lambobin su, da kuma matsayi na 20 mafi karfi kwayoyin ta hanyar tasiri girman.
Bambanci: Ƙarfin faɗakarwa yana da ainihin ginshiƙan fitarwa, ƙofa da buƙatar tabbatarwa. Samfurin yana sarrafa bayanan ku maimakon samar da sunan halittar da aka yi.
uku mini lokuta
Case 1 - Bala'i ba tare da gyara ba: Ƙungiya ta sami 3,800 "muhimmancin" kwayoyin halitta tare da p<0.05 ba tare da gyara ba kuma an gabatar da shi ga bugawa. Lokacin da alkalin wasa ya nemi gyara FDR, jerin sun ragu zuwa kwayoyin halitta 47. Idan da ilimin wucin gadi ya ƙara lambar Benjamini-Hochberg tun daga farko, wannan abin kunya ba zai faru ba. Darasi: gyara ba za a iya tattaunawa ba.
Case 2 - Tasirin Batch: A cikin binciken daya, an sarrafa samfurori a kan kwanaki biyu daban-daban. Abin da suke tunani shine bambancin "haƙuri vs. sarrafawa" shine ainihin "rana ta 1 vs. 2nd day" bambanci (sakamakon tsari: bambancin fasaha saboda ƙungiyar samfur). AI ta taimaka wajen kawar da siginar zato ta hanyar ba da shawarar ƙara madaidaicin tsari zuwa ƙirar (~ tsari + yanayin ƙirar ƙirar).
Hali na 3 - Yin watsi da canjin ninka: Wani ɗalibi ya bayyana "mafi mahimmanci" kwayar halitta wanda bayaninsa ya canza da 2% amma an auna shi da kwanciyar hankali, kawai ta kallon p-darajar. Ganin cewa girman tasirin (log2FC) ya kusan kusan sifili; mahimmancin ƙididdiga ba mahimmancin ilimin halitta ba ne. Samfurin ya bayyana wannan bambancin kuma ya ba da shawarar a gan shi tare da jadawali mai aman wuta.
Teburin kwatance: tsabtar ra'ayi
ra'ayi
Ma'ana
Me yasa yake da mahimmanci?
p-darajar
Yiwuwar bambancin zama kwatsam
kadai zai iya zama mai ɓatarwa
FDR (fadi)
Matsakaicin kuskuren da aka gyara a gwaji da yawa
Yana iyakance abubuwan da ba daidai ba
log2 ninka canji
Girman tasiri
Yana nuna mahimmancin ilimin halitta
tasiri tasiri
Bambancin tsari na fasaha
Yana ƙirƙira siginar karya
al'ada
Gyara ma'auni tsakanin-samfurin
Yana sa kwatanta adalci
Kuskuren gama gari
- Tsallake gyaran kwatancen da yawa: Kuskuren gama gari kuma mafi muni.
- Kawai kallon p-darajar: Tabbatar yin la'akari da girman tasirin (log2FC) tare.
- Ba haɗa da tasirin tsari a cikin ƙirar ba: Kuskure bambancin fasaha don bambancin ilimin halitta.
- Manta al'ada: Kwatanta danyen lambobi kai tsaye.
- Harshen dalili: Cewa "Wannan kwayar halitta tana haifar da cuta"; Bayanan Omics yana nuna alaƙa, dalili yana buƙatar ƙarin gwaji.
Tsanaki: A cikin bayanai masu girma, "mahimmancin ƙididdiga" da "mahimmancin ilimin halitta" abubuwa biyu ne daban-daban. Jerin kwayoyin halittar da basirar wucin gadi ke samarwa shine hasashe na farko; Bai kamata a yi la'akari da kowane jinsin ɗan takara tabbatacce ba tare da tabbatarwa ta hanya mai zaman kanta (qPCR, ma'aunin furotin).
Rage girma da kula da inganci
Abu na farko da za a yi a cikin manyan bayanai shine ganin tsarin tsarin samfurori. PCA (binciken babban bangaren: rage dubunnan masu canji zuwa ƴan gatari kaɗan da nuna su a cikin ma'auni 2) shine daidaitaccen kayan aiki don wannan. Idan ƙungiyoyin da kuke tsammanin (masu sarrafawa/mayya) sun rabu a cikin ginshiƙi na PCA, yana da kyau; amma idan samfuran sun taru ta hanyar "ranar da aka sarrafa" maimakon ta rukuni, wannan gargaɗin tasiri ne. Haka ginshiƙi kuma nan da nan yana nuna misali guda ɗaya (wanda ya gaza).
Zana PCA daga teburin maganata na al'ada (jinin jeri, samfurin shafi). Samfuran launi ta rukuni ( sarrafawa / jiyya), siffar ta hanyar sarrafa tsari. Yi tsokaci kan ko an ga tasirin tsari ko siffa a cikin jadawali.
Wannan matakin heuristic yana motsa sauran binciken: yana da kyau a kama wani abu da wuri fiye da ɓata watanni akan sakamako mai ƙima.
Bayanan salula guda ɗaya: sabon girma
A cikin 'yan shekarun nan, jerin RNA-cell guda ɗaya (sel guda RNA-seq: auna bayanin bayanan dubban ƙwayoyin ɗaiɗai) ya zama tartsatsi. Anan bayanan sun fi girma: dubun duban sel, dubunnan kwayoyin halitta kowanne. Kayan aiki irin su Scanpy (Python) suna sarrafa wannan bayanan; gungun sel kuma yana gano nau'ikan tantanin halitta. AI ya rubuta lambar don wannan aikin aiki, amma ƙirar ilimin halitta na nau'ikan tantanin halitta (ko tari shine "T cell" ko "macrophage") ya dogara da kwayoyin alamomi da ilimin ƙwararru. Tabbatar tabbatar da alamar nau'in tantanin halitta wanda samfurin ya sanya wa gungu tare da sanannun alamomi; Wannan shine matakin da aka fi fahimta a cikin binciken kwayar halitta daya.
Buɗe bayanai da haɓakawa
Yawancin nazarin omics suna loda bayanan su zuwa ma'ajiyar jama'a: GEO (Gene Expression Omnibus) da ArrayExpress don maganganun kwayoyin halitta, SRA (Sequence Read Archive) don raw jerin, PRIDE don proteomics. Wannan yana da mahimmanci don wasu su iya tabbatar da sakamakon ku kuma don ku iya sake nazarin bayanai daga wasu nazarin. AI na iya rubuta lambar (tare da kayan aiki kamar GEOparse) wanda ke saukewa da tsara bayanai daga lambar rajista na GEO (misali lambar GSE); Amma tabbatar da karantawa da tabbatar da ƙirar bayanan da kuka zazzage (rukuni nawa, yawan maimaitawa, wanne tsari) daga rikodin asali. Idan samfurin ya yi iƙirarin "tuna" zane na binciken, wannan kusan ko da yaushe zato ne wanda ke buƙatar tabbatarwa.
a takaice
Bayanan Omics suna auna dubban masu canji a cikin ƙaramin girman samfurin; Wannan yana haifar da tarko na kwatancen da yawa, tasirin batch, da ƙari mai yawa. Hankali na wucin gadi; Yana rubuta lambar don nazarin maganganun banbance-banbance, yana bayyana ra'ayoyin, kuma yana taimaka muku fassara sakamakon. Koyaya, alhakinku ne kuyi amfani da gyaran FDR, kimanta girman tasirin, da kuma guje wa yaren dalili. Kwayoyin halittar ɗan takara su ne hasashe har sai an tabbatar da su ta hanya mai zaman kanta.
Aikin aikace-aikace
Sami ko ƙirƙiri teburin sakamako na banbanci na samfurin (gen, log2FC, padj). Shin AI ta rubuta lambar da ta tace ta FDR<0.05 da |log2FC|>1, ta ba da rahoton adadin mahimman kwayoyin halitta, kuma ya tsara zanen dutsen mai aman wuta. Gudanar da code. Sa'an nan kuma samfurin ya lissafta yawan kwayoyin halitta da za su bayyana "mahimmanci" idan ba a yi gyara ba, kuma ya fassara bambancin.
jerin abubuwan dubawa
- [ ] Na yi amfani da gyaran kwatancen da yawa (FDR).
- Na kimanta girman tasirin (log2FC) da kuma [] p-darajar.
- [ ] Na duba tsari/masu canjin fasaha.
- [ ] Ban tsallake matakin daidaitawa ba.
- [ ] Na yi amfani da harshen haɗin kai maimakon dalili.
- [ ] Na yiwa ɗan takara alama a matsayin hasashe waɗanda ke buƙatar tabbatarwa.