Feeding the Machine: How a Small Lab Generates Big Data

Identifying and quantifying novel compounds of interest is challenging. Utilizing a HTE approach chemists can rapidly synthesize plates in either 384 or 1536 format, creating unique challenges in analysis. Labs need to have the capability to quickly assay any combination of molecules in an untargeted approach with minimal sample consumption to enable downstream assays while reducing human input with a streamlined data workflow. We will demonstrate an end-to-end solution that acquires 2.5nL of crude chemical reaction across 1536 samples in approximately one hour timeframe with minimal human interaction from generation to data processing.  Data is collected from a single 2.5nL droplet ejected from the analytical plate (by commercial Echo Dispenser from Sciex) to a high-resolution mass spectrometer (Sciex 6600 QToF). Acquired spectra are then imported into a prototype processing software that maps each well to its compound of interest with a sonogram representing AUC. Isotopic determination of the molecule is performed via Virscidian, and a readout of all synthesized compounds and their estimated purity are reported and uploaded to LIMS. 52,000 samples were assayed in this format covering both qualitative and quantitative analysis. A single 2.5nL droplet of diluted chemical reaction sample was consumed, leaving the bulk crude for other critical biological assessments. HTE experiment sampling speed does not have as high demand as for HTS screening (samples: thousands vs. millions). For this study 2.4 seconds/injection was utilized to facilitate the downstream sample handling, providing about just over 1 hour sample analysis speed for a full 1536-well nanoscale chemical reaction plate. Data review was simplified via a visual and numerical approach using the following criteria: automatic isotopic pattern matching, PPM error reporting, and peak shape analysis flagging. This was performed on both qualitative and quantitative analysis. For quantitative applications calibration curves were built by linear fitting (1/x weighing setting) of four standards of each compound comprising a subset of 128 compounds. Average coefficient of determination R2 for 128 calibration curves are above 0.998. Crude reactions were analyzed against those standard curves to generate estimated concentrations.