AlphaFold2 e suas aplicações nos campos da biologia e medicina

    Em dezembro de 2020, o AlphaFold2 (AF2), um modelo baseado em aprendizado de máquina (machine learning) para prever estruturas de proteínas desenvolvido pela DeepMind, conquistou o campeonato na 14ª edição do Critical Assessment of Structure Prediction (CASP14). Um ano e meio depois, a DeepMind e o EMBL’s European Bioinformatics Institute (EMBL-EBI) divulgaram as estruturas de mais de 200 milhões de proteínas previstas pelo AF2, que cobrem quase todas as proteínas conhecidas no planeta (o universo das proteínas). Esses dois eventos atraíram grande atenção para o AF2 na comunidade científica. O AF2 representa um avanço marco na previsão de estruturas de proteínas. Ele é considerado a maior contribuição da Inteligência Artificial (IA) para o campo científico e um dos maiores avanços científicos feitos pela humanidade no século XXI. Este é um feito histórico notável no entendimento humano da natureza. A alta avaliação do AF2 não é excessiva, pois entender as estruturas tridimensionais (3D) das proteínas é um dos problemas mais desafiadores no campo da biologia, que tem desconcertado os cientistas por 50 anos. Embora múltiplas tecnologias, incluindo Ressonância Magnética Nuclear (RMN), Cristalografia por Raios-X e Microscopia Crioeletrônica (cryoEM), tenham sido adotadas para resolver as estruturas das proteínas, apenas cerca de 200.000 estruturas de proteínas foram determinadas, cobrindo menos de 0,1% do universo das proteínas.

Adaptado de Yang, Z. et al. (2023). AlphaFold2 and its applications in the fields of biology and medicine. Sig Transduct Target. Tradução livre.

Rain Is Coming to Burning Los Angeles and Will Bring Its Own Risks

    Rain is forecast to begin as soon as Saturday afternoon and to continue as late as Monday evening, says meteorologist Kristan Lund of the National Weather Service’s Los Angeles office. The area desperately needs the precipitation, but experts are warily monitoring the situation because rain poses its own risks in recently burned areas— most notably the potential occurrence of mudslides and similar hazards. “Rain is good because we’ve been so dry,” Lund says. “However, if we get heavier rain rates or we get the thunderstorms, it’s actually a lot more dangerous because you can get debris flows.”

    Fires do a couple of different things to the landscape that can increase the risk of burned material, soil and detritus hurtling out of control. When fires burn hot or long enough, they leave an invisible layer of waxy material just under the surface of the ground. This develops from decomposing leaves and other organic material, which contain naturally hydrophobic or water-repellent compounds. Fire can vaporize this litter, and the resulting gas seeps into the upper soil—where it quickly cools and condenses, forming the slippery layer.

    When rain falls on ground that has been affected by this phenomenon, it can’t sink beyond the hydrophobic layer— so the water flows away, often hauling debris with it. “All of the trees, branches, everything that’s been burned—unfortunately, if it rains, that stuff just floats,” Lund says. “It’s really concerning.” Even a fire that isn’t severe enough to create a hydrophobic layer can still cause debris flows, says Danielle Touma, a climate scientist at the University of Texas at Austin. Under normal conditions, trees and other plants usually trap some rain above the surface, slowing the water’s downward journey. But on freshly burned land there’s much less greenery to interfere; all the rain immediately hits the ground. [...]

    Fortunately, the rain should also help firefighters tame the blazes that remain active. The largest, the Palisades Fire, is currently 77 percent contained. The second largest, the Eaton Fire, is 95 percent contained. The Hughes Fire is third largest and only 56 percent contained. A fire can be fully contained but still burning. The containment percentage refers to the amount of the perimeter that has barriers that firefighters expect will prevent further spread.

Scientific American. January 27th, 2025. Adaptado.