Naúrar 2 / 11

Wakilin Kwayoyin Halitta da Tsarin Sinadarai: Tabbatarwa tare da SMILES, InChi da RDKit

Riba:

  • Ikon tantance kwayoyin halitta ba da suna ba amma ta tsarin wakilcin (SMILES tare da mutane, InChI/InChIKey tare da bayanan bayanai)
  • Ƙarfin gano ingantattun sifofi ko kuskure ta hanyar ƙididdigewa, ingantawa da kuma daidaita kowane SMILES da aka bayar ta hanyar hankali na wucin gadi tare da RDKit.
  • Samun damar fahimtar cewa ƙididdige ƙididdiga kamar nauyin kwayoyin halitta da dabara ya kamata a samo su daga kayan aiki na tushen ƙa'ida, ba daga ƙirar harshe ba.

Sharadi na farko don amfani da AI cikin aminci a cikin ilmin sunadarai shine rubuta kwayoyin halitta a cikin yaren da na'ura da dan adam zasu iya fahimta. Kuna ce "phenol", AI yana samun daidai; amma idan ka ce "acid" akwai dubban yiwuwar. Sunayen suna da shubuha; tsarin wakilci daidai ne. A cikin wannan rukunin, za mu koyi mahimman bayanai guda biyu waɗanda ke fassara kwayar halitta zuwa rubutu (SMILES da InChi) da kayan aikin da ke tantance su (RDKit). Manufarmu: don ba da kwayoyin halitta ga AI ta hanyar da ba za ta taba fahimta ba, kuma don tabbatar da tsarin da AI ta samar tare da kayan aiki.

Menene MURMUSHI?

SMILES (Simplified Molecular-Input Line-Entry System) sigar rubuta kwayoyin halitta a cikin layi daya na rubutu. Atom ana wakilta ta haruffa, shaidu ta alamomi, da zobe ta lambobi. Misalai:

  • Ethanol: CCO (carbon-carbon-oxygen; hydrogens a fakaice).
  • Benzene: c1ccccc1 (ƙananan "c" yana nuna carbon aromatic; 1's rufe zobe).
  • Aspirin: CC(=O)Oc1ccccc1C(=O)O.
  • Caffeine: Cn1cnc2c1c (= O) n (C) c (= O) n2C.

Ikon SMILES shine cewa yana ɗaukar tsarin haɗin gwiwa gaba ɗaya a cikin layi ɗaya; Rashin rauninsa shine cewa kwayoyin halitta guda ɗaya na iya samun MURMUSHI masu inganci masu yawa (wannan ana kiransa non-canonicality). Za mu magance wannan tare da RDKit a cikin ɗan lokaci.

Alamomi: "SIMILES na canonical" na kwayoyin halitta shine guda ɗaya, daidaitaccen rubutun waɗancan kwayoyin halitta. Don ganin ko MURMUSHI guda biyu kwayoyin halitta iri ɗaya ne, gwada su kuma kwatanta su; Kada su zama daidai a rubutu, amma su kasance daidai da kwayoyin halitta.

Menene InChi?

InChi (International Chemical Identifier) ​​daidaitaccen mai ganowa ne wanda IUPAC ya haɓaka. Bambanci daga MURMUSHI shine cewa kwayoyin guda ɗaya koyaushe suna ba da InChi iri ɗaya; wato canonical a cikin yanayi. Yana da tsawo kuma yana da wuyar karantawa, amma ya dace don daidaita bayanai. Ana amfani da gajeriyar “InChIKey” (halaye 27) don nema.

  • Aspirin InChIKey: BSYNRYMUTXBXSQ-UHFFFAOYSA-N.

Doka: Mutane suna magana SMILES (karanta), inji da ma'ajin bayanai sun dace da InChi/InChIKey (daidai). Ba da murmushi ga AI; Tabbatar a cikin bayanan bayanai tare da InChiKey.

RDKit: injin tabbatar da kayyade

RDKit babban ɗakin karatu ne na cheminformatics. Ba kamar ƙirar yare ba, tushen ƙa'ida ne: ƙaddamar da SMILES, ƙididdige nauyin kwayoyin halitta, haifar da SMILES na canonical, dawo da InChi/InChIKey, zana kwayoyin. Don haka RDKit "yana ƙididdige" abin da AI "annabta". Wannan shine zuciyar tafiyar aiki: AI ke haifarwa, RDKit ya tabbatar.

Ainihin kwararar tabbatarwa:

daga rdkit shigo da Chemfrom rdkit.Chem shigo da Bayani, Drawsmiles = "CC(=O)Oc1ccccc1C(=O)O" Chem.MolToSmiles(mol)) buga ("Molecular formula:", Chem.rdMolDescriptors.CalcMolFormula(mol)) printers ("MolFormula(mol)) madaurin nauyi ("MolMol) 2), "g/mol") bugawa ("InChIKey:", Chem.MolToInchiKey(mol))

Idan MolFromSmiles ya dawo Babu, AI ya ba ku ƙwayoyin cuta mara inganci; Wannan kadai gargadi ne mai kima. Idan inganci, ba kwa buƙatar ƙara tambayar ƙirar harshe don nauyin kwayoyin halitta; Yana a hannunku ƙudiri.

Mataki-mataki: haɗin gwiwar AI + RDKit

  1. Gano kwayoyin halitta: Ba AI sunan, nemi MURMUSHI. Misali, "tabbatar da SMILES na canonical don paracetamol."
  2. Tabbatar: Sanya MURMUSHI mai shigowa tare da MolFromSmiles. Idan Babu, ƙi.
  3. Canonicalize: Dawo da rubutun canonical tare da MolToSmiles; Koyaushe amfani da wannan daga yanzu.
  4. Lissafin lambobi: Sami nauyin kwayoyin halitta, dabara, adadin zobe, adadin masu ba da gudummawa/masu karɓa da sauransu daga RDKit, ba daga AI ba.
  5. Gyara ID: Ƙirƙirar InChIKey, bincika a cikin bayanan bayanai (PubChem), tabbatar da cewa kwayar halitta ita ce mahadi da kuke so.

Samfura huɗu masu kwafi

1) Neman SMILES (daga suna zuwa tsari):

Aiki: Ba da SMILES na canonical don mahallin mai zuwa. Hadarin: paracetamol (acetaminophen). Doka: Ba da kirtani SMILES kawai da InChIKey, kar a ƙara wani ƙarin bayani. Idan ba ku da tabbas, rubuta "USURE", kar a gyara.

2) Buƙatun tabbatar da SMILES (daga tsari zuwa sarrafawa):

Yi nazarin MURMUSHI da ke ƙasa: CC (=O) Nc1ccc (O) cc1Tambayoyi:1) Shin wannan MURMUSHI mai inganci?2) Menene fili yayi daidai da (sunan gama gari)?3) Menene ma'anar kwayoyin halitta? Lura: Zan lissafta ainihin ma'aunin kwayoyin tare da RDKit; Kuna ba da fassarar tsarin ku kawai.

3) Kwatanta wakilci biyu:

Ina da MURMUSHI guda biyu:A) OCCB) CCOTask: Waɗannan kwayoyin halitta iri ɗaya ne ko daban? Rubuta dalilin ku. Lura: Zan binciko amsar ku ta hanyar canza ta a cikin RDKit.

4) Bayanin tsari (don dalilai na koyo):

MURMUSHI: Cn1cnc2c1c(=O)n(C)c(=O)n2CTtask: Karanta wannan SMILES yanki guda kuma ka bayyana ma’anar kowace alama (atom, bond, zobe, aromaticity) a fili cikin Baturke. Kuma gaya wace ce mahadi.

Rauni mai ƙarfi / Ƙarfi mai ƙarfi

Rauni:

Ba da kaddarorin acetaminophen.

"Feature" yana da m; AI yana haifar da lambobi bazuwar daga nauyin kwayoyin halitta zuwa wurin narkewa, wasu daga cikinsu zasu yi kuskure.

Mai ƙarfi:

Compound: acetaminophen.1) Canonical SMILES da InChIKey ver.2) Ba da tsarin kwayoyin halitta.Dokar: KADA KA BA DA NUMERIC dabi'u kamar nauyin kwayoyin halitta, narkewa, da dai sauransu; Zan same su da RDKit/PubChem. Kawai ba da ID ɗin gini.

Bambanci: Mun jagoranci AI zuwa ga abin da yake da karfi a (tsari na ainihi) kuma daga abin da yake da rauni a (lambobin gwaji).

Kwatanta abubuwan gani

fasali

MURMUSHI

InChi / InChiKey

Iya karantawa

Babban (abokan abokantaka)

Low (mai son na'ura)

canonicity

Batun canzawa (yana buƙatar canonization)

ta halitta marar aure

Amfani

sadarwar mutum, zane, shigarwa

Database match, ainihi

stereochemistry

Taimako (@ alamomin)

Taimako (mai layi)

don ba AI

manufa

Mafi dacewa don tabbatarwa

kananan lokuta

Case 1 - Sunaye biyu, kwayoyin halitta daya. Wani dalibi ya yi tunanin cewa "N-acetyl-para-aminophenol" da "paracetamol" sun kasance mahadi daban-daban kuma sun tsara gwaje-gwaje daban-daban guda biyu. Lokacin da aka tsara MURMUSHInsu a cikin RDKit, dukkansu sun zama CC (= O) Nc1ccc (O) cc1; Kwayoyin halitta iri ɗaya ne. Lost: rabin yini na shiri; riba: za a iya kauce masa tare da duban canonicalization na minti 2.

Hali na 2 - Kama MURMUSHI mara inganci. AI ta mayar da CC(=O)Nc1ccc(O)cc1C(don wani bambance-bambancen (brackets ba a rufe ba).Dalibin ya gudu MolFromSmiles, ya dawo Babu, nan da nan ya ga kuskuren kuma ya nemi gyara SMILES. Ba tare da tabbatarwa ba, tsarin da ba daidai ba zai iya yada zuwa duk asusun.

Case na 3 - Nauyin kwayoyin da ba daidai ba. YZ ya ce "nauyin kwayoyin 214.3 g/mol" don ibuprofen (C13H18O2). Lissafin RDKit ya ba da 206.28 g/mol. Bambanci don 0.1 mol: 21.43 g tare da YZ, ainihin 20.63 g. Wannan bambanci na 3.9% zai rushe stoichiometry da lissafin yawan amfanin ƙasa. Ya kawo lissafin ƙididdiga.

Kuskuren gama gari

  • Bayar da kwayoyin halitta da suna. Synonyms/sunayen ciniki sun ruɗe. Koyaushe haɗa SMILES ko InChi.
  • Kwatanta MURMUSHI ba tare da canonicalize shi ba. OCC da CCO sun bambanta a rubutu amma iri ɗaya ne a cikin kwayoyin halitta.
  • Tambayar nauyin kwayoyin halitta zuwa samfurin harshe. Wannan shi ne lissafin ƙaddara; Ana yin shi daga RDKit ko da hannu daga dabarar.
  • Rashin lura da MURMUSHI marasa inganci. Rashin duba dawowar MolFromSmiles Babu kuma ci gaba da tsarin da bai dace ba.
  • Rashin kula da ilimin kimiya na zamani. Enantiomers biyu (masu isomers na madubi) na iya nuna tasirin ilimin halitta daban-daban; Tsallake alamar @/@ @ a cikin MURMUSHI.
Hankali: Tsarin kwayoyin halitta iri daya baya nufin kwayoyin halitta daban-daban. C2H6O shine duka ethanol (CCO) da dimethyl ether (COC). Daidaiton dabara ba daidaiton ainihi ba ne; Yi amfani da InChiKey don ganewa.

A takaice

  • Gano kwayoyin halitta ta hanyar bayanin tsari, ba da suna ba: SMILES tare da mutum, InChI/InChIKey tare da bayanan bayanai.
  • RDKit yana da ƙaddara; AI yana annabta, RDKit yana ƙididdigewa kuma yana tabbatarwa.
  • Fassara kowane AI MURMUSHI tare da MolFromSmiles kuma bincika Babu, sannan canza tsarin.
  • Sami lambobi kamar nauyin kwayoyin halitta da dabara daga RDKit, ba daga ƙirar harshe ba.
  • Daidaiton tsari ba yana nufin ainihin kwayoyin halitta ba; Yi amfani da InChiKey don ganewa.

Aikin aikace-aikace

Zaɓi mahadi guda uku (misali maganin kafeyin, ibuprofen, glycine). Tambayi AI don Canonical SMILES ga kowane. Sannan rubuta rubutun RDKit (amfani da samfurin da ke sama) kuma samar da: Canonical SMILES, dabarar kwayoyin halitta, nauyin kwayoyin halitta, InChIKey. Nawa ne daga cikin MURMUSHI da AI suka bayar sun kasance masu inganci? Idan YZ kuma ya ba da nauyin kwayoyin halitta, ƙididdige bambanci azaman kashi tare da ƙimar RDKit. Shirya bincikenku a cikin ƙaramin tebur.

jerin abubuwan dubawa

  • [ ] Na san menene SMILES da InChi/InChIKey da lokacin amfani da su.
  • [ ] Na tabbatar da duk MURMUSHI da AI ke bayarwa tare da MolFromSmiles.
  • [ ] Na canonicalize SMILES kafin in kwatanta su.
  • [ ] Ina samun nauyin kwayoyin halitta da dabara daga RDKit, ba daga samfurin harshe ba.
  • [ ] Na tabbatar da asalin kwayoyin halitta a cikin bayanan tare da InChIKey.
  • [ ] Na san yanayi inda stereochemistry ke da muhimmanci.