Riba:
- Fahimtar kulawar inganci, daidaitawa, kiran bambance-bambancen da matakan bayanin matakan bincike kuma ku sami damar yin amfani da hankali na wucin gadi cikin aminci a cikin rubutun da taƙaita sakamakon.
- Ikon tabbatarwa da kuma fitar da kwayoyin halitta, sunadarai, da nassoshi ta hanyar haɗa kowane fassarar jeri zuwa ma'auni kamar kashi na ainihi, ɗaukar hoto, da ƙimar e-daraja.
- Ikon fassara hasashen samfurin harshen furotin a matsayin yiwuwar, tabbatar da ƴan takara masu mahimmanci tare da gwajin rigar, da kuma amfani da horo na raba bincike daga ganewar asibiti.
Mafi mahimmancin kayan albarkatun bioengineering shine jerin (jeri - jerin wakilcin DNA, RNA ko furotin da aka rubuta a cikin haruffa; misali ATGCGT ... ko MKV don furotin ...). Na'urar jerawa tana samar da ɗaruruwan miliyoyin gajerun karatu na dare; Ayyukan bioinformatics ne (dabi'ar da ke nazarin bayanan ilimin halitta tare da hanyoyin lissafi) don canza wannan ɗanyen bayanai zuwa amsawar ilimin halitta mai ma'ana. A cikin wannan rukunin, zaku koyi inda zaku yi amfani da AI cikin aminci a cikin bincike na jeri, waɗanne daidaitattun kayan aikin zasu hanzarta shi, kuma inda hukuncin ƙwararrun ku yake da makawa.
Matakai na yau da kullun a cikin bincike na jeri sune: kula da ingancin karatun ɗanyen karatu (cire karantawa mara kyau da ragowar adaftar), daidaitawa zuwa ga kwayar halitta (daidaitawar - gano inda karatun ya fito daga kwayar halitta), kiran bambance-bambancen (gano maye gurbi, maki inda mutum ya bambanta da ma'anar), da fassarar binciken. AI yana taimakawa a kowace hanyar haɗi a cikin wannan sarkar, ko dai ta hanyar rubuta rubutun, taƙaita sakamakon, ko samar da daftarin sharhi; amma matakai masu mahimmanci kamar daidaitawa da kiran bambance-bambancen har yanzu ana yin su tare da ingantattun kayan aikin yau da kullun.
Daidaita jerin: kamanni da ma'ana
Aunawa irin kamannin jeri biyu shine zuciyar bioinformatics. BLAST (Kayan aikin Neman Daidaita Gida na asali - kayan aiki na yau da kullun wanda ke kwatanta jeri zuwa jeri a cikin manyan ma'ajin bayanai kuma ya sami mafi kamanceceniya) shine matakin farko na fahimtar menene furotin ko kwayar halitta. Bambance ra'ayoyi guda biyu a cikin jeri jeri: ainihi (yawan jeri biyu suna da harafi ɗaya) da e-daraja (ƙididdigar da ke nuna yuwuwar kamanni da aka samu ta faru kwatsam; idan ƙarami ne, yana da mahimmanci). AI na iya fassara fitowar BLAST kuma ya ba da jita-jita kamar "wannan jeri yana da yuwuwar kinase enzyme," amma kun tabbatar da wannan fassarar tare da ƙimar e-daraja, ɗaukar hoto, da sanannun bayanan yanki.
Tukwici: Lokacin tambayar AI don kewayon tsokaci, koyaushe nemi ma'aunin daidaitawa da kyau: ainihin kashi, ɗaukar hoto, e-darajar. Idan ba tare da waɗannan ma'auni ba, fassarar da aka bayar na "wannan sunadaran shine" ba za a iya tabbatar da shi ba kuma yana iya zama hallucination.
Tsarin tsararru na tushen AI
A cikin 'yan shekarun nan, nau'ikan yaren furotin waɗanda ke ɗaukar jeri kamar "harshe" (samfurin AI waɗanda aka horar da miliyoyin jerin sunadaran sunadaran kuma suna hasashen abubuwan aiki da tsarin tsarin jeri; kamar ESM) sun bayyana. Waɗannan na iya yin hasashen ko maye gurbi zai rushe furotin, wanda dangi jerin ya kasance, ko yankuna masu aiki. Yana da kayan aiki mai ƙarfi mai ƙarfi: yana ware dubunnan bambance-bambancen kafin gwaji kuma yana ba da haske ga waɗanda suka fi dacewa. Amma tsinkaya shine yiwuwar; Ana gwada kowane ɗan takara mai mahimmanci ta hanyar gwaji.
Tsanaki: Lokacin da samfurin harshen furotin ya ce "wannan maye gurbi yana da lalacewa", wannan ƙima ce mai yiwuwa, ba ganewar asali ba. Ana ba da fassarar asibiti (misali rahoton bambance-bambancen cuta) tare da ingantattun kayan aiki, ƙayyadaddun ƙayyadaddun kayan aikin da ƙwararrun shawarwarin ƙwayoyin cuta.
uku mini lokuta
Case 1 - An ƙara ƙarin bayani. A cikin aikin metagenomics ɗaya, ƙungiyar ta ci karo da jerin furotin 8,400 waɗanda ba a san su ba daga samfuran muhalli. Bayanin farko na taimakon AI (bayyana alamun ayyuka zuwa jeri) ya tara su zuwa iyalai masu aiki kuma sun samar da taswirar farko cikin awanni 6. Ƙungiyar kawai ta karɓi babban amincewa 30%; da hannu ya tabbatar da ragowar tare da BLAST da HMM.
Case 2 - Halucination kama. Wani dalibi ya tambayi AI "wace kwayoyin halitta jerin jerin ya fito da kuma tushen DOI." AI ta ba da ainihin sunan nau'in da bayanin labarin. ɗalibin ya sanya shi a kan BLAST: wasan da ya fi kusa ya kasance aji daban gaba ɗaya tare da ID na 41% kuma babu tunani. AI ta samar da cikakkiyar amsa amma gyara.
Case 3 - Bambance-bambancen tacewa. Ɗaya daga cikin aikin kwayoyin halitta yana da bambance-bambancen danyen mai miliyan 4.7. AI ya rubuta rubutun tacewa wanda ya haɗa da inganci, zurfin, da madaidaicin yawan yawan jama'a; ƙasa zuwa 120 bambance-bambancen da suka dace na asibiti. Ƙungiyar ta ba da tabbacin kowane tacewa tare da labarin tushen kuma ta gudanar da rubutun da kansa; Sakamakon ya zama mai yiwuwa.
Samfura huɗu masu kwafi
1) Rubutun sarrafa ingancin FASTQ:
Matsayinku: injiniyan bioinformatics. Rubuta rubutun a Python: shigarwa fayil ne na FASTQ, fitarwa shine matsakaicin ingancin karantawa, abun ciki na GC, da ragowar adaftar. Yi amfani da Biopython azaman ɗakin karatu. Bayyana lambar kuma rubuta abin da kowace ƙimar kofa (misali Q30) ke nufi. Daidaita aikin da babu shi.
2) Bayanin fitarwa na BLAST (dangane da tushe):
Zan ba ku fitowar tambura ta BLAST (ginshiƙai: tambaya, batu, % ainihi, alignment_length, kimantawa, bitscore). Rubuta ID, iyaka da e-darajar magana ga kowane bugun. Ƙidaya kawai waɗanda ke da e-daraja <1e-5 da ɗaukar hoto> 70% a matsayin "amintattu". Gina fassarar ku akan waɗannan ma'auni; Alama nau'in/aiki hasashen a matsayin "yana buƙatar tabbaci".
3) Bambancin dabaru na tacewa:
Matsayinku: mai binciken bioinformatician ba na asibiti ba. Ba da shawarar matakan tacewa don VCF: ƙaramin zurfin karantawa, ƙimar inganci, iyakar yawan jama'a. Faɗa dalili da tushen kowane ƙofa. Wannan tace bincike ne; Rubuta a kan bugun cewa ba za a iya amfani da shi don ganewar asibiti ba.
4) Bayanin ingantawa na Codon:
Ta yaya zan inganta amfani da codon yayin zayyana jerin DNA don bayyana jerin furotin don [kwayoyin halitta]? Bayyana matakai da kasada don yin la'akari (ma'auni na GC, maimaita jeri, wuraren ƙuntatawa). Faɗa mani waɗanne kayan aikin tabbatarwa da zan yi amfani da su kafin samar da tsararrun kankare.
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rawanin faɗakarwa:
Yi nazarin wannan jeri: ATGCGTACGT...
Wani nau'i na bincike, wane ma'auni, wane sakamako ba shi da tabbas; AI yana samar da fassarori kyauta waɗanda ke buɗe don ƙirƙira.
Ƙarfi mai ƙarfi:
Matsayinku: injiniyan bioinformatics. Don jerin DNA masu zuwa tare da Python/Biopython: (1) ƙididdige tsayi da abun ciki na GC, (2) nemo buɗaɗɗen firam ɗin karatu (ORFs) a cikin firam ɗin karatu shida, (3) fassarar furotin na ORF mafi tsayi. Ba da rahoton sakamako daga lamba kawai; yin fassarar aikin biological.Series: [jerin]
Bambanci: bayyanannen ƙananan ayyuka, daidaitaccen kayan aiki, ingantaccen fitarwa dangane da lamba, da iyakar sharhi.
Ayyukan bincike na jeri da kuma rawar AI
nema
daidaitaccen abin hawa
AI gudunmawa
tabbatarwa
kula da inganci
FastQC, Trimmomatic
Rubutun + taƙaitawa
Madaidaicin ma'auni
daidaitawa
BWA, Bowtie2
Shawarar siga
Ikon rabo na daidaitawa
Bambancin kira
GATK, bcftools
Daftarin aikin
An tabbatar da bambance-bambancen da aka sani
annotation
BLAST, InterProScan
Pre-clustering
E-darajar + yanki
Ƙimar tasiri
ESM, SIFT
Matsayin ɗan takara
rigar gwaji
Kuskuren gama gari
- Karɓar sharhi ba tare da awo ba. Ba tare da kashi ɗaya ba, ɗaukar hoto, da ƙimar e-daraja, cewa "wannan furotin ne" ba za a iya tabbatarwa ba.
- Rikita tsarin tunani / daidaitawa. Idan nau'in genome (hg38 vs hg19) da 0/1 na tushen daidaitawa sun haɗu, bambance-bambancen wurare ba za su yi daidai ba.
- Yin watsi da tasirin tsari. Samfuran da aka jera a batches daban-daban suna jagorantar mutum zuwa kuskuren bambance-bambancen fasaha don ilimin halitta.
- Yin amfani da sunan aikin AI da aka yi. Samfurin na iya haifar da sunayen kwayoyin halitta / furotin / kayan aikin da ba su wanzu; Tabbatar da kowane suna a kan ainihin bayanan bayanai.
- Ba da fassarar asibiti ta hanyar kayan aikin bincike. Ƙungiyar bambance-bambancen tare da cutar ana ba da rahoto ne kawai ta hanyar yarda, ƙayyadaddun matakai.
a takaice
Binciken jeri shine albarkatun halittu na bioengineering, kuma AI yana haɓaka kowane mataki na wannan sarkar ta hanyar rubutawa, taƙaita fitarwa, da manyan ƴan takara. Amma matakai masu mahimmanci kamar daidaitawa, kira daban-daban, da aikin aikin ana yin su tare da daidaitattun kayan aiki, ingantattun kayan aiki, kuma kowane sharhi yana ɗaure da ma'auni kamar adadin ID, ƙimar e-daraja, da tabbatarwa. Samfuran harshen furotin kayan aikin tantancewa ne masu ƙarfi; Fitowar su abu ne mai yuwuwa, ba ƙarewa ba har sai an tabbatar da su ta hanyar gwaji.
Aikin aikace-aikace
Zaɓi jerin samfurin da ake samuwa a bainar jama'a (misali jinsin halitta daga kwayar halitta). Ka sa AI ta fara yin bincike na asali na asali (abun ciki na GC, ORF, fassarar) tare da samfurin "ƙarfi mai ƙarfi". Sannan tabbatar da abin da aka fitar da kansa tare da Biopython ko kayan aikin kan layi kuma lura da kowane bambance-bambance. A ƙarshe, tambayi AI tambayar sharhi da ke neman tushe, kuma bincika cewa duk wani nassoshi da ya bayar daidai ne ta hanyar duba su cikin ainihin bayanan.
jerin abubuwan dubawa
- [ ] Na tabbatar da kowane sharhin kirtani tare da ma'auni na ainihi / ɗaukar hoto / e-darajar.
- [ ] Na fayyace sigar genome da tsarin daidaitawa.
- [ ] Na gudanar da bambance-bambancen rubutun nazari da kansa kuma na sake buga shi.
- [ ] Na fassara hasashen samfurin furotin a matsayin yiwuwar, ba sakamako ba.
- [ ] Na tabbatar da duk gene/protein/ sunayen sunaye da AI suka bayar game da ainihin bayanan.
- [ ] Na raba bincike na bincike daga ganewar asibiti.