Expert-level protocol translation for self-driving labs

TL;DR

A framework that automates the protocol translation process through a three-stage workflow, which incrementally constructs Protocol Dependence Graphs (PDGs) that approach structured on the syntax level, completed on the semantics level, and linked on the execution level is proposed.

Abstract

Recent development in Artificial Intelligence (AI) models has propelled their application in scientific discovery, but the validation and exploration of these discoveries require subsequent empirical experimentation. The concept of self-driving laboratories promises to automate and thus boost the experimental process following AI-driven discoveries. However, the transition of experimental protocols, originally crafted for human comprehension, into formats interpretable by machines presents significant challenges, which, within the context of specific expert domain, encompass the necessity for structured as opposed to natural language, the imperative for explicit rather than tacit knowledge, and the preservation of causality and consistency throughout protocol steps. Presently, the task of protocol translation predominantly requires the manual and labor-intensive involvement of domain experts and information technology specialists, rendering the process time-intensive. To address these issues, we propose a framework that automates the protocol translation process through a three-stage workflow, which incrementally constructs Protocol Dependence Graphs (PDGs) that approach structured on the syntax level, completed on the semantics level, and linked on the execution level. Quantitative and qualitative evaluations have demonstrated its performance at par with that of human experts, underscoring its potential to significantly expedite and democratize the process of scientific discovery by elevating the automation capabilities within self-driving laboratories.

Figures (5)

• Figure 1: Illustration of the protocol translation problem. An nl-based protocol is translated to a structured protocol, then to a completed protocol, and finally to a linked protocol that is ready for self-driving laboratories along with a corresponding pdg, after being processed through the syntax, semantics, and execution levels. The three colors of arrows and text/ code highlights indicate the three translation steps respectively.
• Figure 2: The design principles and the resulting pipeline of our translator.(Syntax level) Operation dependence synthesis on the syntax level, through the joint optimization of dsl program syntax space and the parsing tree of the nl-based protocols. This process is static and context-free. (Semantics level) Reagent flow analysis on the semantics level, through an automaton scheme maintaining the lifecycles of reagents and intermediate products. This process is static and context-free. (Execution level) Spatial-temporal dynamics analysis on the execution level, through the partial execution trace model based on the spatial-temporal dual constraint representation. This process is dynamic and context-aware.
• Figure 3: Results of experiment.(A) Distinctions between various domains regarding domain-specific corpora and the corresponding dsl. (B) Convergence of the three indicators in the objective function for program synthesis. (C) Our translator significantly outperforms the best baseline and approaches human-level performance. (D) Our translator significantly outperforms alternative methods on the syntax level. (E) Our translator significantly outperforms alternative methods on the semantics level.
• Figure 4: Showcases of the results.(A) Examples of the final pdg generated by our translator, the alternative method, and human experts. (B) Examples of structured protocols output by our translator and alternative methods. (C) Examples of completed protocols output by our translator and alternative methods.
• Figure : Reagent flow analysis