跳到正文
insitro· @insitro · X·· 11 天前AI 评分35
AI 导读

insitro 发布 Deep Phenotype 第三期《Getting Data to Grow in a Dish》,回顾 POSH 平台构建生物数据工厂的进展。细胞与基因扰动匹配率从 34% 升至 71% 以上,新细胞分割模型开发周期从 6–12 个月缩短到约 1 个月。重点是用自动化、AI 和人类专业能力生成可靠、可复现的数据,服务 AI 驱动药物发现。

正文 · 原文

Out now: “Getting Data to Grow in a Dish,” the third issue of Deep Phenotype, insitro’s Substack on scaled biology and deep causality.

Generating biological data at scale is one challenge. Ensuring that data is reliable, reproducible and meaningful enough to power AI-driven drug discovery is another.

In our third installment, insitro founder and CEO @DaphneKoller and insitro director of integrated technology exploration, @ATGmsalickTAG, take readers behind the scenes of building insitro's biological data factory, where automation, AI, and human expertise come together to generate high-quality data and deepen our understanding of disease biology.

Through the evolution of our Pooled Optical Screening in Human cells (POSH) platform, Daphne explores the extensive scientific and engineering advances required to turn complex cellular experiments into scalable, reproducible workflows.

Along the way, the share of cells we could match to their genetic perturbation more than doubled, from 34% to above 71%, and developing a new cell-segmentation model went from six to twelve months to about a month.

“The most important thing we have built is not any single platform. It is the ability to build the next one.”
╰┈➤ Read the latest installment of Deep Phenotype: https://deepphenotype.substack.com/p/getting-data-to-grow-in-a-dish
#DrugDiscovery #AI #Biology #CausalBiology

来源:insitro · x.com