Riba:
- Samun damar fahimtar aikin AlphaFold, ƙididdiga masu aminci kamar pLDDT da PAE, kuma cewa tsarin shine 'hasashen', da fassara tsarin daidai da hankali na wucin gadi.
- Ikon gane wuraren da ke da ƙananan ƙididdiga masu aminci (mai sassauƙa / rashin daidaituwa) da kuma guje wa fassarar wuce gona da iri da kwatanta tare da shaidar gwaji
- Ikon yin amfani da horo na lura da tsinkayar tsari azaman hasashe da haɗa da'awar aiki zuwa tabbacin gwaji.
Don fahimtar abin da furotin ke yi, sau da yawa ya zama dole a san tsarinsa mai ninke mai girma uku; saboda aiki yana tasowa daga tsari. Shekaru da yawa, tantance tsarin furotin ta gwaji (X-ray crystallography, cryo-electron microscopy) zai ɗauki watanni zuwa shekaru. AlphaFold (tsarin hankali na wucin gadi wanda DeepMind ya haɓaka wanda ke yin hasashen tsarin furotin daga jerin amino acid) ya rage wannan zuwa mintuna kuma ya sanya sifofin da aka annabta na ɗaruruwan miliyoyin sunadaran sunadaran jama'a. A cikin wannan rukunin za ku koyi yadda ake karanta fitowar AlphaFold, yadda ake fassara makin ƙarfin gwiwa, da yadda ake amfani da LLM cikin aminci cikin wannan fassarar.
Bambanci mai mahimmanci: AlphaFold kayan aikin tsinkayar tsari ne kuma fitowar sa tsinkaya ce, ba gaskiyar gwaji ba. LLM (samfurin taɗi kamar ChatGPT) ba AlphaFold ba ne; Ba zai iya "tunawa" haɗin haɗin furotin ba. Yi amfani da LLM don fassara da bayyana fitowar AlphaFold; dawo da tsarin kanta daga ma'ajin bayanai na AlphaFold ko kayan aiki.
Ka'idoji na asali
- Tsarin farko: Amino acid jerin (sarkar wasiƙa na furotin).
- Tsarin Sakandare/Tsaro: Helix (alpha-helix), takardar (beta-sheet) da nadawa sarkar mai girma uku.
- pLDDT: Ƙididdiga na gida na AlphaFold ga kowane amino acid, wanda ke jere daga 0 zuwa 100. 90+ yana nufin babban ƙarfin gwiwa, 70-90 yana nufin mai kyau, 50-70 yana nufin ƙasa kaɗan, kuma ƙasa da 50 yana nufin ƙaramin ƙarfi. Yankunan ƙananan pLDDT gabaɗaya yankuna ne na “rashin tsari” (ba tare da ninki daban ba).
- PAE (Kuskure Haɗin Kai): Kuskuren da aka tsinkaya a cikin matsayi na yankuna biyu; Ya nuna yadda dogara ga jeri na yankuna ya danganta da juna.
- PDB: Database da tsarin fayil wanda a cikinsa ake adana tsarin furotin na gwaji.
Karatun fitowar AlphaFold: mataki-mataki
1. Zaɓi madaidaicin furotin da jeri. Gano furotin tare da lambar UniProt (bayanin bayanan sunadaran); Nemo tsarin da wannan lambar a cikin ma'ajin bayanai na AlphaFold.
2. Dubi taswirar pLDDT. Dubi waɗanne sassa na tsarin an annabta tare da babban tabbaci (blue) kuma waɗanda aka annabta tare da ƙarancin amincewa (orange / rawaya). Yi hankali game da sharhi a wuraren da ba a amince da su ba.
3. Karanta ginshiƙi na PAE. PAE yana nuna ko tsarin dangi na yanki a cikin furotin mai yawan yanki abin dogaro ne. Babban PAE yana nufin cewa matsayin dangi na filayen ba shi da tabbas.
4. Kwatanta tare da tsarin gwaji. Daidaita tsarin gwaji a cikin PDB idan akwai. Idan hasashen AlphaFold yana kusa da gwaji, kwarin gwiwar ku yana ƙaruwa.
5. Fassara bambancin tasirin. Lokacin fassara tasirin canjin amino acid akan tsarin, la'akari da pLDDT da mahimmancin aikin wannan yanki tare (wuri mai aiki, aljihun ɗaure).
Tukwici: Ƙananan pLDDT ba koyaushe yana nufin "rashin zato" ba; Yawancin lokaci alama ce da ke nuna cewa yankin yana da matsala a zahiri. Don haka ƙarancin amincewa wani lokaci yana nuna ingantaccen gaskiyar halitta. Hakanan duba jerin tare da kayan aikin tsinkayar rashin bin ka'ida don bambanta wannan.
uku mini lokuta
Shari'a 1 - Ƙarfafa fassarar yankin rashin amincewa. Ɗaya daga cikin ɗalibi ya yi iƙirarin cewa "madauki" a cikin tsarin AlphaFold ya dace a cikin wani aljihu, kuma AI ta tabbatar da hakan. Duk da haka, lokacin da ɗalibin ya kalli pLDDT, ya ga cewa makin ɗin ɗin ya kasance 42 (ƙananan sosai); don haka matsayinsa bai kasance abin dogaro ba. An janye da'awar. Darasi: Koyaushe duba makin amana na gida kafin yin tsokaci.
Case 2 - Haɗin kai da aka yi. Wani mai bincike ya tambayi LLM haɗin 3D na wani amino acid na furotin; LLM ya ba da lambobi masu gamsarwa zuwa wurare goma sha uku. Lokacin da mai binciken ya kwatanta su da ainihin haɗin kai a cikin fayil ɗin AlphaFold PDB, ya ga cewa an ƙirƙira su gaba ɗaya. LLM ba zai iya "tunawa" haɗin gwiwar ba; dole ne a karanta masu daidaitawa daga fayil ɗin.
Case 3 - Tabbatar da samu. Wani dalibin PhD yayi mamakin ko bambance-bambancen daya (p.Gly12Val) yana kusa da wurin aiki na furotin. YZ ta tunatar da cewa an ƙayyade ragowar rukunin yanar gizon a cikin UniProt; Dalibin ya buɗe UniProt kuma ya sami wurin aiki, sannan auna shi tare da mai duba tsarin (PyMOL) cewa Gly12 ya kasance 8 angstroms daga wannan rukunin yanar gizon a cikin tsarin AlphaFold. Ma'aunin lamba ya ƙunshi da'awar "kusa".
Misali: karanta pLDDT daga tsarin fayil
# A cikin fayil ɗin AlphaFold PDB, ana adana pLDDT a cikin ginshiƙi na B. daga Bio.PDB shigo da PDBParserimport numpy as npyapi = PDBParser(QUIET=Gaskiya).get_structure("AF", "AF-P04637-F1.pdb")plddt = [atom.bfactor don zarra a tsarin.get_atoms() idan zarra.name == "Average"]D zagaye (np.ma'ana (plddt), 1)) bugawa ("Ƙarfin amincewa (<70) ƙimar saura (%):", zagaye (100 * np.ma'ana (np.array (plddt) <70), 1))
Wannan yana ba ku damar ganin lambobi gaba ɗaya amincin tsarin kafin yin sharhi.
Samfura huɗu masu kwafi
1) Fassarar tsari (tare da ƙima):
Matsayinku: mataimakiyar ilimin halitta. Taimaka min fassara tsarin AlphaFold na furotin UniProt [lamba]. Amsa waɗannan tambayoyin amma daidaitawa/maki FITTING: waɗanne yankuna muke tsammanin babban / low pLDDT, menene PAE ke nunawa, akwai tsarin gwaji (PDB), ta yaya zan kwatanta? Zan karanta dabi'u daga fayil ɗin.
2) Tasirin bambancin-tsari:
Taimaka min fassara yiwuwar tasirin bambance-bambancen mai zuwa akan tsarin furotin: [p. Ka ba ni shaidar da zan nema maimakon iƙirarin “hallaka” kai tsaye.
3) bayanin pLDDT/PAE:
Yi bayani da misali bambanci tsakanin pLDDT da PAE, wanne ya kamata in duba da kuma lokacin da ke cikin bambance-bambancen bincike. Yi la'akari da shari'o'in furotin na yanki guda da na yanki daban daban.
4) Tsarin kwatanta tsari:
Ina so in kwatanta hasashen AlphaFold tare da tsarin PDB na gwaji. Yadda za a lissafta RMSD (ma'anar sabawa tsakanin sifofi), wane kayan aiki zan yi da (PyMOL, Biopython), menene madaidaicin ƙidaya a matsayin "mai kyau zoba"? Ba da tsarin mataki-mataki.
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rauni: "Zana tsarin wannan sunadaran kuma ku faɗi tasirin p.Arg273His."
Matsala: LLM ba zai iya zana tsarin daidaitawa / dacewa ba; ba a la'akari da maki amincewa; Da'awar tabbataccen tasiri ba zai zama marar tushe ba.
Mai ƙarfi: "Na sauke AlphaFold ginawa don UniProt P04637 da kaina. Duban pLDDT na p.Arg273Sauran sa da bayanin aikin sa a cikin UniProt, gaya mani mataki-mataki yadda zan fassara tasirinsa; ba ni irin shaidar da zan nema maimakon tabbataccen da'awar."
Dalilin da ya sa yake da ƙarfi: Ana ɗaukar tsarin daga tushe, an sanya maƙasudin amana a cibiyar, da'awar tana da alaƙa da shaida.
Tambaya
Shin AlphaFold yana bayarwa?
Inda / yadda ake tabbatarwa
Hasashen ninka 3D
Ee
AlphaFold DB / fayil
amanar gida
Da (plDDT)
B-factor shafi
Wuri na yanki
Da (PAE)
Farashin PAE
Tsarin gwaji na ainihi
a'a
PDB (gwaji)
Daidaitaccen tasirin aikin bambance-bambancen
a'a
Nazarin aiki + gwani
Kuskuren gama gari
- Tsallake makin amincewa. Fassarar a cikin ƙananan yankin pLDDT ba abin dogaro bane.
- Kuskure hasashe don haƙiƙanin gaskiya. Fitowar AlphaFold kimanta ce; Hukunce-hukuncen yanke shawara suna buƙatar tabbataccen shaida.
- Neman haɗin kai daga LLM. Ana karanta haɗin kai daga fayil ɗin, ba a haddace ba.
- Yin watsi da PAE. A cikin sunadaran multidomain, matsayin dangi na yankuna na iya zama mara tabbas.
- Nan da nan la'akari da ƙarancin amincewa a matsayin "kuskure". Wani lokaci wannan yanki ba daidai ba ne.
Tsanaki: AlphaFold yana gina hoton "a tsaye"; Ka tuna cewa sunadaran haƙiƙan ƙwayoyin ƙwayoyin hannu ne waɗanda zasu iya shigar da ƙima. Babu wani tsari guda ɗaya da ke nuna cikakkiyar ɗabi'a mai ƙarfi.
Zurfin: Inda AlphaFold yayi shuru cikin tsari
AlphaFold yana da ƙarfi, amma sanin abin da ba zai iya yi ba shine rabin yaƙin amfani da shi lafiya. Na farko, daidaitaccen AlphaFold yana ninka furotin guda ɗaya a sararin samaniya; Ba ya yin samfurin ligands, ions, cofactors da kwayoyin kwayoyi. Zinc ion a cikin wurin aiki na enzyme ko substrate ba ya bayyana a cikin tsarin; Saboda haka, sharhi kamar "wannan aljihun babu komai, rashin aiki" na iya zama yaudara. Na biyu, ba kai tsaye ya ƙirƙiri tasirin maye gurbin ba: ko da kun gabatar da bambance-bambancen guda biyu kusa da jeri ɗaya, AlphaFold yawanci yana samar da kusan tsari iri ɗaya; Ba za ku iya tabbatar da da'awar "wannan bambance-bambancen yana karya tsari" ta kwatanta tsinkaya biyu daga AlphaFold. Na uku, baya la'akari da gyare-gyaren bayan fassarorin (kamar phosphorylation, glycosylation) da ainihin yanayin sunadaran membrane a cikin membrane.
Harka a cikin batu: ƙungiya ɗaya ta yi iƙirarin daga hoton AlphaFold cewa bambance-bambancen kusa da aljihun ɗaurin ATP a cikin enzyme kinase "ya rufe aljihu"; ilimin wucin gadi kuma ya tabbatar. Lokacin da aka buɗe tsarin kristal na gwaji (PDB), ya bayyana cewa mai haɗin gwiwa a wannan yanki da AlphaFold bai ƙirƙira ya riga ya tsara aljihu ba - tasirin bambance-bambancen ya kasance daga wata hanya ta daban. Al’amari na biyu: wani mai bincike ya yi hasashen cewa “madauki” tsakanin fage guda biyu ya hada fage biyu; Taswirar PAE ta nuna babban kuskure (matsayin dangi mara kyau) tsakanin waɗannan yankuna biyu, don haka da'awar haɗin ba ta da tushe. Karatun PAE ya hana labari mara kyau.
5) Samfurin sarrafa iyakokin AlphaFold:
Kafin yin la'akari da da'awar tsari mai zuwa, jera waɗanne iyakoki na AlphaFold suka shiga cikin wannan tambayar: [da'awa]. Faɗa mini abin da ƙarin shaida (tsarin gwaji, nazarin aiki) Ina buƙata, musamman dangane da rashi na ligand / ion / cofactor, rashin ƙirar maye gurbi, da rashin tabbas na PAE.
A takaice
- AlphaFold kayan aiki ne mai ƙarfi wanda ke hasashen tsari daga jerin abubuwa, amma fitowar sa hasashe ne; ya kamata a karanta tare da amintaccen maki.
- pLDDT yana nuna amana na gida, PAE yana nuna amintaccen wurin yanki; Dubi duka biyu kafin yin sharhi.
- LLM ba zai iya "tunawa" daidaitawa ko tsari ba; sami tsarin daga AlphaFold/PDB, yi amfani da LLM a cikin sharhi.
- Shaida ta zahiri da kuma ƙwararrun ƙwararrun ƙwararrun ƙwararrun ƙwararrun ƙwararrun ƙudiri ne.
Aikin aikace-aikace
Zazzage tsarin AlphaFold na furotin (misali, ingantaccen bincike mai hana ƙari) ta lambar UniProt. Yi ƙididdige matsakaicin plDDT da ƙarancin amintaccen ragowar rabo tare da snippet lambar da ke sama. Sannan zaɓi bambance-bambancen kuma nemi AI don tsarin fassarar tare da samfuri na 2; Bincika bayanin pLDDT da UniProt na wannan saura da kanka. Ɗaure da'awar ku zuwa ƙimar amincewar lamba.
jerin abubuwan dubawa
- [ ] Na sami tsarin daga AlphaFold/PDB tare da lambar UniProt.
- [ ] Na karanta pLDDT da PAE kafin yin sharhi.
- [ ] Ban nemi LLM don yin daidaitawa/maki ba.
- [ ] Na danganta bambance-bambancen fassarar zuwa ma'aunin amincewa na ragowar abin da ya dace.
- [ ] Na kwatanta shi da tsarin gwaji, idan akwai.
- [ ] Na goyi bayan yanke shawara mai mahimmanci tare da ƙwararrun ƙwararrun ƙwararrun ƙwararrun ƙwararrun shaida.