Loading market data...

EPFL Researchers Build AI That Picks Optimal Molecular Synthesis Routes from Plain-Language Commands

EPFL Researchers Build AI That Picks Optimal Molecular Synthesis Routes from Plain-Language Commands

. Translate headings. Let's write the translated content step by step. First, the opening paragraph: "Researchers at the Swiss Federal Institute of Technology Lausanne (EPFL) have developed an artificial intelligence framework that can take a chemist’s everyday language instructions and pick the best way to make a given molecule from thousands of possible synthesis paths. The system, which the team described in a recent paper, aims to cut down the time chemists spend manually searching through reaction databases and planning multi-step routes." Translation: "スイス連邦工科大学ローザンヌ校(EPFL)の研究者らは、化学者が日常的に使う言語の指示を受け取り、数千もの可能な合成経路の中から特定の分子を製造する最適な方法を選択できる人工知能フレームワークを開発した。チームが最近の論文で発表したこのシステムは、化学者が手作業で反応データベースを検索し、多段階の経路を計画する時間を削減することを目的としている。" Second paragraph: "How the AI works" heading: "AIの仕組み" Then paragraph: "The framework uses natural language processing to understand commands like “make this compound starting from benzene” or “find a route that avoids toxic solvents.” It then runs a search through known chemical reactions and molecular structures to find the most efficient sequence of steps. The researchers said the system considers factors such as yield, cost, and safety when ranking the options." Translation: "このフレームワークは自然言語処理を使用して、「この化合物をベンゼンから合成する」や「有毒な溶媒を避ける経路を見つける」といったコマンドを理解する。そして、既知の化学反応と分子構造を検索し、最も効率的な一連のステップを見つけ出す。研究者らによると、システムはオプションをランク付けする際に収率、コスト、安全性などの要素を考慮する。" Next paragraph: "Unlike earlier tools that required users to input exact SMILES strings or reaction templates, this one accepts the kind of free-text instructions a chemist would write in a lab notebook. That lowers the barrier to using computational planning, especially for researchers who aren’t programming experts." Translation: "以前のツールではユーザーが正確なSMILES文字列や反応テンプレートを入力する必要があったのに対し、このツールは化学者が実験ノートに書くような自由形式のテキスト指示を受け付ける。これにより、特にプログラミングの専門家でない研究者にとって、計算による計画立案のハードルが低くなる。" Next heading: "Potential applications in chemistry" -> "化学における潜在的な応用" Paragraph: "The ability to quickly map out a synthesis route matters most in fields like pharmaceutical development, where getting a new drug candidate made in the lab can be a bottleneck. A search that might take a human hours or days can be compressed into minutes by the AI." Translation: "合成経路を迅速に計画できる能力は、医薬品開発のような分野で最も重要であり、そこでは新しい医薬品候補を実験室で合成することがボトルネックとなることがある。人間なら数時間から数日かかる検索が、AIによって数分に短縮できる。" Next paragraph: "The EPFL team tested the framework on a set of target molecules that included common pharmaceutical intermediates. In each case, the system returned a plausible route that matched or improved on the one a human chemist had designed. The researchers noted that the AI sometimes suggested a different sequence of reactions that was shorter or used cheaper reagents." Translation: "EPFLチームは、一般的な医薬品中間体を含む一連の標的分子でこのフレームワークをテストした。いずれの場合も、システムは人間の化学者が設計したものと一致するか、それを改善する妥当な経路を返した。研究者らは、AIがより短い、またはより安価な試薬を使用する異なる反応シーケンスを提案することもあると指摘した。" Next heading: "What the researchers accomplished" -> "研究者の成果" Paragraph (first one): "The work was done by a group in EPFL’s Laboratory of Computational Chemistry and Organic Synthesis. They trained the AI on a large database of known chemical reactions, giving it the ability to break down a complex target into simpler building blocks. The system then works backward — a strategy called retrosynthesis — to find the steps." Translation: "この研究は、EPFLの計算化学・有機合成研究室のグループによって行われた。彼らは既