Scientists’ Side Hustle? Using AI and Quantum Computing to Generate New Peptides

科学家的“副业”？利用人工智能与量子计算生成新型肽

Scientists have successfully shown a quantum computer can improve the accuracy and reach of generative artificial intelligence drug discovery models. And they did it using their spare time and money leftover from other projects. 科学家们已成功证明，量子计算机能够提高生成式人工智能药物发现模型的准确性和覆盖范围。而这一切，竟是他们利用业余时间和从其他项目中节省下来的经费完成的。

The Technical University of Denmark team ran their generative AI model for predicting proteins in conjunction with a printer-sized quantum computer built by British startup ORCA Computing, which sped up AI by linking quantum machines with traditional processors. The researchers used the hybrid technique to generate novel peptides—short chains of amino acids—capable of binding to specific proteins in the body. Doing so is a crucial step in vaccine development. 丹麦技术大学（DTU）的研究团队将他们用于蛋白质预测的生成式 AI 模型，与英国初创公司 ORCA Computing 开发的一台打印机大小的量子计算机相结合，通过将量子机器与传统处理器连接，加速了 AI 的运行。研究人员利用这种混合技术生成了新型肽（氨基酸短链），这些肽能够与体内的特定蛋白质结合。这是疫苗开发中的关键一步。

The team of researchers worked weekends and pooled unspent money from other projects because “most innovative science is too scary for foundations,” according to DTU professor Timothy Patrick Jenkins, who led the project. 据该项目负责人、DTU 教授 Timothy Patrick Jenkins 称，研究团队利用周末时间工作，并汇集了其他项目中未用完的资金，因为“大多数创新科学对基金会来说都太‘可怕’（风险太大）了”。

Making the peptides in the laboratory and testing whether these would bind to the particular proteins showed the model produced more successful peptides than its classical counterpart, with the strongest improvements where training data was rare. 在实验室中制造这些肽并测试它们是否能与特定蛋白质结合，结果显示该模型比传统模型产生了更多成功的肽，尤其是在训练数据稀缺的情况下，改进效果最为显著。

The team believe the machine could accelerate the development of personalized immunotherapies and vaccines, as well as improve drugs’ efficacy in understudied groups. 该团队认为，这种机器可以加速个性化免疫疗法和疫苗的开发，并提高药物在研究不足群体中的疗效。

“We needed to really prove it to convince skeptics that our predictions connect to the real world,” Patrick Jenkins tells WIRED. Quantum computing remains a nascent field and faces intense scrutiny due to the technical challenges of building these machines and successfully applying them to solve problems. “我们需要真正证明这一点，以说服怀疑论者相信我们的预测与现实世界是相连的，”Patrick Jenkins 告诉《连线》（WIRED）杂志。量子计算仍是一个新兴领域，由于在构建这些机器以及成功将其应用于解决问题方面存在技术挑战，该领域正面临严格的审视。

Even Patrick Jenkins was initially reluctant to explore the technology: “I was a huge quantum skeptic” he says with a laugh, believing any application to his work would be “decades away.” 就连 Patrick Jenkins 最初也不愿探索这项技术：“我曾是一个坚定的量子怀疑论者，”他笑着说，认为任何将其应用于自己工作的尝试都还要“几十年”才能实现。

He and his team use big data and AI to discover proteins which could unlock new immunotherapies cheaper and faster, often funded by the Novo Nordisk Foundation. While most biological model makers are desperate for more data, a particular challenge for his team has been the lack of data on the full variety of genetic information across the human race, since most medical research has focused on Western populations. This can make it difficult to develop peptides that will work on understudied populations, such as those in Asia and Africa, he says. 他和他的团队利用大数据和人工智能来发现蛋白质，从而以更低成本、更快速地解锁新的免疫疗法，这些研究通常由诺和诺德基金会（Novo Nordisk Foundation）资助。虽然大多数生物模型构建者都渴望获得更多数据，但该团队面临的一个特殊挑战是缺乏全人类遗传信息的多样性数据，因为大多数医学研究都集中在西方人群身上。他说，这使得开发适用于亚洲和非洲等研究不足人群的肽变得困难。

His team hypothesized that embedding a quantum computer into their workflow could make it generate a more diverse set of peptides, especially for targets where they had less data, after learning that the machines had a similar effect in generating images. 在了解到量子计算机在生成图像方面具有类似效果后，他的团队假设，将量子计算机嵌入工作流程中，可以使其生成更多样化的肽，特别是在数据较少的靶点上。

The newly discovered process won’t revolutionize research yet as quantum computers are still too small to run full-scale, cutting-edge AI models, meaning better results could be achieved on a classical computer. 这一新发现的过程目前还不会彻底改变研究现状，因为量子计算机规模仍然太小，无法运行全规模、尖端的 AI 模型，这意味着在传统计算机上仍能获得更好的结果。

“Quantum is still not very powerful, so the level of complexity that we could encode wasn’t a normal-sized antibody, which is what we usually work with,” says DTU PhD student Jonathan Funk. Furthermore, finding a peptide that can bind to a specific gene is just one step in vaccine development, and wouldn’t alone yield successful drugs. “量子技术目前还不够强大，因此我们能编码的复杂程度还达不到我们通常处理的正常大小抗体的水平，”DTU 博士生 Jonathan Funk 说。此外，找到能与特定基因结合的肽只是疫苗开发的一步，单凭这一点并不能直接产生成功的药物。

“I think it’s no surprise that lots of industrial companies think quantum is hazy and far away,” ORCA Computing chief executive officer Richard Murray tells WIRED, partly because the technology “has not ever had really clear near-term examples of usefulness.” “我认为很多工业公司认为量子技术模糊且遥远并不奇怪，”ORCA Computing 首席执行官 Richard Murray 告诉《连线》，部分原因是这项技术“从未有过真正明确的近期实用案例”。

He says this study is novel in that it shows a near-term commercial application for quantum. His company is also applying the technology through projects with oil major BP on chemistry and carmaker Toyota on making its design process more efficient. 他表示，这项研究的创新之处在于它展示了量子技术在近期的商业应用。他的公司也正在通过与石油巨头英国石油公司（BP）在化学领域的合作，以及与丰田汽车在提高设计流程效率方面的合作，来应用这项技术。

The DTU team will now see if it can use the workflow with more cutting-edge models and larger proteins. “We needed this as an easy way to validate that now we actually have a shot at moving the needle substantially,” says Patrick Jenkins, noting that generative AI workflows are particularly valuable in neglected diseases that receive little research money. He’s also looking at using a quantum computer to enhance his generative AI method for designing synthetic antidotes for snakebite venom. DTU 团队现在将研究是否能将该工作流程应用于更尖端的模型和更大的蛋白质。“我们需要通过这种简单的方式来验证，我们现在确实有机会取得实质性的进展，”Patrick Jenkins 说。他指出，生成式 AI 工作流程在那些研究经费匮乏的被忽视疾病领域尤为宝贵。他目前还在考虑利用量子计算机来增强其生成式 AI 方法，以设计针对蛇毒的合成解毒剂。