Introduction: Umlinganiselo Olandelayo ye-AI ku-Medicine Uma thwebula i-AI emzimbeni, sincoma ngokuvamile ukuthi isebenze umsebenzi yokubonisa imibala ye-supernatural; ukucacisa izidakamizwa zonyango kwezidakamizwa ezincinane noma ukucubungula ama-datasets amancane ukucacisa imiphumela yama-patient. Lezi zokusebenza ezinzima, kodwa kubandakanya kakhulu ukufumana ukuxhumana kwebhizinisi esiyingqayizivele. Lokhu kubaluleke umbuzo enhle kakhulu: kungase-AI ungaphakeme ukucacisa okufakelwa kwama-akhawunti etholakalayo ukuze zibonele ngokuvamile izifundo zenzululwazi? Ukubuyiselwa I-C2S-Scale yenzelwe, imodeli ye-27 billion parameter foundation eyenziwe ku-Google's Gemma family of open models. Lesi-AI yenzelwe akuyona kuphela ukuhlola idatha ye-cellular, kodwa nge-motivation ngokuvamile ukukhiqiza i-hypothesis entsha, enokutholakalayo yokwelapha umdlavuza. Umhlahlandlela wahlala, ukuhlola umzila we-biological ebonakalayo ngaphambi kokubili ukuthi kungasiza izinhlelo zethu zokusiza ukwelashwa kwama-tumors ezithile. Lezi ziphumela kunezinguquko eyodwa; inikeza indlela entsha yokwenza i-AI inokusebenza njenge-partner yokudala ku-research yenzululululwazi, ukwakha indlela yokwenza i-hypothesis eyenziwe nge-AI. Lapha izindlela ezine eziphambili ezivela kule ukuhlaziywa. Okokuqala, umkhumbi we-AI uqeqeshiwe ukufunda ulimi lwe-Life Ukulungiswa okuhlobene iyisisombululo enhle ebizwa ngokuthi "Cell2Sentence" (C2S). Ngokuvamile, le ndlela ibheka idatha yokudluliselwa kwe-gen ephikisana kusuka ku-cell eyodwa kuya ku-format ukuthi I-Large Language Model (LLM) ingathola: isihloko. Lokhu "isihloko se-cell" isakhiwa kusuka ku-genes e-K eyenziwe kakhulu, eyenziwe ngokuhambisana ne-level yayo yokudluliselwa. Kuyinto enhle ngoba ivumela abacwaningi ukusetshenziswa i-stat-of-the-art LLMs, eyakhelwe ngokuvamile ukucubungula ingoma womuntu, ngqo idatha eziphilayo emangalisayo. Ngaphandle kokucubungula izakhiwo ze-AI ezintsha, ezivamile ze-biology, i-C2S i-reformates inkinga yemvelo ukuze ifakwe isixhobo esiyingqayizivele enhle. Lokhu kususa izinzuzo eziphambili kanye nezinhlelo zokusebenza zokusebenza ze-LLMs for i-single-cell analysis. It Found A Therapy That Works Smarter, Akukho Harder Umthamo omkhulu ekwelapheni umdlavuza wokwelapha umdlavuza kuyinto yokuba "ukudluliselwa" umdlavuza-umdlavuza okuyinto ngempumelelo engatholakali uhlelo lwesimo. Umthamo ebalulekile we-immunotherapy kuyinto ukwenza lezi umdlavuza "ukudluliselwa" ngokuzimela ukubonisa ama-immuno-triggering nge-process ebizwa ngokuthi i-antigen presentation. I-antigen engaphezu kwe-cancer cell ibonise, kunzima kakhulu ku-immuno cells okuyinto zihlanganisa. I-C2S-Scale I-I.I.I. inikeza umsebenzi ephelele kakhulu usebenzisa inqubo enhle ebizwa ngokuthi i-"dual-context virtual screen." Akungabikho kuphela i-drug enikeza ukubonisa i-antigen yonke isikhathi; bayibiza i-model ukufumana "i-amplifier e-conditional." Ukuze ukwenza lokhu, i-AI i-analyzed amasu amabili: a "immuno-context-positive" setting usebenzisa amamakhasimende amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu amakhulu Kwangathi okokuqala. Waze Ngokwenza lokhu, kulindeleke izinga le-conditional reasoning eyenziwe ngempumelelo le-27-billion-parameter model, njengoba amamodeli amancane asikwazi ukuguqulwa lokhu imiphumela esihlalweni. Ngemva kokuguquka imiphumela ye-4200 kwezidakamizwa, i-AI ibonise i-kinase inhibitor ebizwa ngokuthi i- silmitasertib (CX-4945). I-AI yenza i-hypothesis entsha kakhulu Umphumela we-I.I. yokubuyekeza wama-novelty yayo. I-modeli awukwazi nje ukucacisa ukuxhumana okufakiwe kumadokhumenti yayo yokwakhiwa. Ukuxhumana okufakiwe phakathi kwe- silmitasertib (CX-4945) ne-antigene yokubuyekeza ekusebenziseni kwe-interferon akuyona engaphansi kwebhizinisi lwezenzululululwazi. I-AI yenza umqondo olungcono ngokuphelele. Lokhu kubonisa ukuhlaziywa okuphambili kusuka ku-pattern recognition kuya ku-hypothesis yokuqala, ukwandisa I-AI ku-realm of a genuine research partner. Njengoba abacwaningi wabelane embhedeni yabo: "Ngaphandle kokuba i-CK2 iye ifakwe ngezinto eziningi ze-cellular, kuhlanganise njenge-modulator ye-immunity, ukuchithwa kwe-CK2 nge-silkitasertib ayikho kwebhizinisi yokukhangisa ngokuvamile ukuxhumana kwe-MHC-I noma ukucubungula kwe-antigen. Lokhu kubonisa ukuthi imodeli yenza i-hypothesis entsha enokutholakalayo, futhi akuyona kuphela imiphumela ebonakalayo." Ukusuka ku-Digital Prediction kuya ku-Lab-Verified Reality Ukubuyekezwa kwe-AI, kungekho okuhlobisa, kuyinto kuphela iphefumulo kuze kube luhlolwa. Isinyathelo esifundeni esifundeni kuyinto ukuthatha isicelo se-model. (i-computer-based) ukubuyekeza futhi ukubuyekeza Umhlahlandlela wahlola i-hypothesis kumamodeli ye-neuroendocrine yabantu (okufinyelelwa kwamaseli we-Merkel ne-pulmonary origins) - i-cell types that were minimally represented in the model's training data, okwenza i-validation ngaphezulu kakhulu. Ukusuka Silicon In vitro Iziphumo zokusebenza kwezilebhu zithiphile ukuhlaziywa kwe-AI ngokucacileyo: Ukwelashwa kwamakhemikhali nge-drug silmitasertib kuphela ayikho imiphumela ku-antigen presentation. Ukwelashwa kwamakhemikhali nge-dose encane ye-interferon kuphela kuneziphumela enhle. Ukwelashwa kwamakhemikhali nge-silkitasertib kanye ne-interferon e-low-dose, njengoba i-AI wabheka, ikhiqiza "ukukhanyisa okukhanyisa, synergistic." Ngokuvamile, ukwelashwa kwamakhemikhali kunikeza ukwanda kwe-50% ku-antigen presentation. Lo mphumela ungayenza amangqamuzana amangqamuzana amangqamuzana kakhulu emzimbeni emzimbeni, ukulawula i-AI entsha hypothesis kanye nokuvumela isitimela esizayo esizayo yokwelashwa kwamakhemikhali. I-Blueprint entsha ye-Discovery Umphumela wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo wesixazululo. Ngokuyinhloko, amabhizinisi we-Google ne-Yale akuyona kuphela ukubukeka kwabo, kodwa futhi isixhobo esebenzayo. I-C2S-Scale model kanye nezinsizakalo ziye zithunyelwe kumaziko wesayensi, okukhuthaza amanye abacwaningi ukwakha kule umsebenzi. Lokhu kuthatha nathi ukubukeka: Uma i-AI iyahambisana nathi ukufundisa ulimi oluthile yamaseli ethu, ukuthi ezinye izinguquko zebhizinisi zithunyelwe nathi? Ukukhanya kwe-podcast Ukukhanya kwe-podcast Apple: lapha Spotify: lapha Ngiya Ngiya
