nanoNET

The Anthropocene era is marked by the increasing dispersion of nanoparticles (NPs) in the environment, whether natural or man-made. Due to their small size and high reactivity, these NPs play a key role in the transport of trace elements and their biogeochemical transformations, requiring advanced analytical tools to study them. Single-particle inductively coupled plasma time-of-flight mass spectrometry (spICP-TOFMS) has recently emerged as a reference method, offering real-time multi-element characterization of individual NPs. This project proposes to integrate artificial intelligence approaches, notably sequence-to-sequence (Seq2Seq) models, to effectively exploit this massive data and improve the detection and identification of NPs in spICP-TOFMS time series. The objective is to develop an optimized analytical pipeline that can: i/ accurately detect NP peaks despite signal variability, ii/ automatically identify NP families based on their elemental composition, and iii/ standardize these methods for broader application in environmental nanogeochemistry, which generates large volumes of data. To achieve this, we will compare several Seq2Seq models and use advanced classification and clustering algorithms (DBScan, Spectral clustering), first using simulated time series to develop the methodology, which we will then validate using spICP-TOFMS data from the nanoSTREAM project (CNRS EC2CO).