Machine learning assisted construct design to accelerate protein production.

学术界的药物发现和生物医学研究在很大程度上依赖于蛋白质的生产，例如生物测定、新型活性化合物的高通量筛选和蛋白质晶体结构的测定。然而，在药物发现项目启动过程中，以足够的质量和数量生产和交付特定蛋白质往往是第一步，也是限速的一步。蛋白质结构目前是根据经验丰富的科学家的专业知识手工设计的，需要反复的试验和错误优化，这是一个劳动密集型的过程。因此，该项目的重点将是准确预测其组成氨基酸序列的蛋白质产量。为了实现这一目标，在基因组研究方面具有丰富经验的生物信息学家Dilrini De Silva博士将被借调到阿斯利康，为期六个月。她将在阿斯利康的定量生物学（数据科学系）和蛋白质生产团队的交叉点工作。

Drug discovery and biomedical research in academia relies heavily on the production of proteins for example in bioassays, high-throughput screening for novel active compounds and the determination of protein crystal structures. _However, producing and delivering a specific protein to partners in sufficient quality and quantity is often the first and rate-limiting step in the initiation of drug discovery projects._Protein constructs are currently being designed manually based on the expertise of experienced scientists and require repeated rounds of trial and error optimisations, making this a labour-intensive process. Therefore, the focus of this project will be to accurately predict protein production yield from its constituent amino acid sequences.To achieve this goal Dr. Dilrini De Silva, a bioinformatician with extensive experience in genomic research will be seconded to AstraZeneca for a period of six months. She will be operating at the intersection of Quantitative Biology (a data science department) and Protein Production teams within AstraZeneca.