Purifying the masses: integrating prepurification quality control, high-throughput LC/MS purification, and compound plating to feed high-throughput screening

Machine learning models and performance dependency on 2D chemical descriptor space for retention time prediction of pharmaceuticals

The predictive modeling of liquid chromatography methods can be an invaluable asset, potentially saving countless hours of labor while also reducing solvent consumption and waste. Tasks such as physicochemical screening and preliminary method screening systems where large amounts of chromatography data are collected from fast and routine operations are particularly well suited for both leveraging large datasets and benefiting from predictive models. Therefore, the generation of predictive models for retention time is an active area of development. However, for these predictive models to gain acceptance, researchers first must have confidence in model performance and the computational cost of building them should be minimal. In this study, a simple and cost-effective workflow for the development of machine learning models to predict retention time using only Molecular Operating Environment 2D descriptors as input for support vector regression is developed. Furthermore, we investigated the relative performance of models based on molecular descriptor space by utilizing uniform manifold approximation and projection and clustering with Gaussian mixture models to identify chemically distinct clusters. Results outlined herein demonstrate that local models trained on clusters in chemical space perform equivalently when compared to models trained on all data. Through 10-fold cross-validation on a comprehensive set containing 67,950 of our company's proprietary analytes, these models achieved coefficients of determination of 0.84 and 3 % error in terms of retention time. This promising statistical significance is found to translate from cross-validation to prospective prediction on an external test set of pharmaceutically relevant analytes. The observed equivalency of global and local modeling of large datasets is retained with METLIN's SMRT dataset, thereby confirming the wider applicability of the developed machine learning workflows for global models.

Ultrasonic sample introduction combined with flame assisted thermal ionization: Pretreatment-free direct mass spectrometry analysis for fraction collecting tubes of preparative liquid chromatography

Ultrasonic sample introduction combined with flame assisted thermal ionization mass spectrometry (USI-FATI-MS) was developed to monitor the fractions of preparative liquid chromatography. Recently, ultrasound-based sample introduction techniques have achieved great advance in the field of high-throughput analysis. However, it is still a challenge to directly apply these existing techniques to the analysis of macro volume samples (mL level). In this work, ultrasonic sample introduction combined with flame assisted thermal ionization was used for pretreatment-free direct mass spectrometry analysis of micro to macro volume samples (μL-mL level). Utilizing this unique design of ultrasonic sample introduction, liquid sample in the container can be quickly atomized to the gas phase without contact. Then, due to the flame assisted thermal ionization source, desolvation and ionization of the sample droplets will occur immediately. USI-FATI-MS has shown excellent sensitivity, repeatability and great compatibility to solvents and compounds with a wide range of polarity. As a proof of concept, USI-FATI-MS has been applied for rapid monitoring and identification of purified synthetic and natural products in fractions.

Just-in-Time Purification: An Effective Solution for Cherry-Picking and Purifying Active Compounds from Large Legacy Libraries

Many companies possess a compound collection consisting of purified compounds and of unpurified products from combinatorial libraries. Using commercial and proprietary compounds as examples, this report provides clear examples of the significant impact purification can have on the activity observed for a compound and highlights the need to retest the purified compounds prior to creating structure-activity relationships. Crude mixtures made with commercial compounds led to an increase in the number of false positives in the SXR-GAL4 assay as compared with their pure and purified counterparts. An examination of proprietary compounds in an HIV assay resulted in the purification of 61 active crude synthetic mixtures. Of these 61 compounds, 32 were 5-fold less active and 2 were 5-fold more active after purification. This report details a semiautomated process developed and implemented for cherry-picking, tracking, and selectively purifying compounds found active in high-throughput screening campaigns.

High-throughput purification platform in support of drug discovery

The application of parallel synthesis is an efficient approach to explore the chemical space and to rapidly develop meaningful structure activity relationship (SAR) data for drug discovery programs. However, the effectiveness of the parallel synthesis requires a high throughput purification workflow that can process a large number of crude samples within a meaningful time frame. This paper describes a high throughput purification platform that has been adopted at Merck's Rahway research site. The platform includes the evaluation of crude samples, mass-directed HPLC purification, fraction analysis, compound registration, final compound purity assessment and assay distribution. Assisting with the sample tracking and the data management is the internally designed laboratory information management system, Light Automation Framework (LAF). Using this process and the tools described herein, the group has successfully achieved purities of 95% or greater for 90% of samples.

Advances in improving the quality and flexibility of compound management

The process of drug discovery has evolved considerably since the advent of high-throughput screening (HTS) in the 1980s. Experts and opinion leaders today are agreeing that the current trend in the field is a focus on increasing overall quality (target, screening, and compounds), use of multiple screening approaches for lead discovery, and more flexibility in the process. The associated need for increased flexibility and quality control to support existing HTS paradigms as well as lower throughput approaches such as fragment screening, computational chemistry, focused library building, and centralized lead optimization support has required an evolution in compound management (CM, aka sample management or library management). Although there is much less published peer-reviewed data in CM, due to its historical links to HTS, it has followed very similar trends. In recent years, the focus in CM has been increasingly in compound quality and increased flexibility of the process, as opposed to number of compounds dispensed and speed of dispensing, which were standard metrics and indicators used not so long ago. Ideally, to screen the highest quality sample for every assay, one would start with a correct identity and pure solid, make a correct concentration solution in water or water-soluble/assay-compatible solvent that would allow 100% solubilization, and screen it immediately in a biological assay. Neither CM nor screening has advanced sufficiently to deliver this ideal scenario, but many significant advancements have been made in recent years both in terms of quality of compounds in stores and flexibility of the process, which will be reviewed herein.

Automated medicinal chemistry

With the advent of high throughput technologies in biological screening in the 1980s, providing sufficient numbers of small molecules for screening became a bottleneck in the drug discovery process. Combinatorial chemistry was the first attempt by chemists to address this issue. However, since its first applications, combinatorial chemistry has evolved rapidly into diverse fields. This review will focus on the evolution and the current status of what we refer to today as automated medicinal chemistry.

A systematic investigation of recovery in preparative reverse phase high performance liquid chromatography/mass spectrometry

In this paper we report a systematic recovery study based on reversed phase high performance liquid chromatography (RP-HPLC) separation and mass spectrometric (MS) based fractionation. Factors including a compound's physicochemical properties, column mass loading and presence of impurities were investigated through commercially available compounds. Results suggest that the delay time between MS peak detection and fraction collection, fraction detector's signal-to-noise ratio and compound's base peak width in the chromatogram have the biggest impacts on purification recovery. In an effort to assess sample recovery within our high throughput purification process, re-purification was performed on four compound libraries that were synthesized in-house. Reproducible recoveries (>80%) were achieved in all tests.

Cherry-Picking in an Orchard: Unattended LC/MS Analysis from an Autosampler with >32,000 Samples Online

The 1536-well microplate format has widely supplanted the 384-well microplate format for high-throughput screening and for IC50 assays. Previously, liquid chromatography/mass spectrometry (LC/MS) analyses of such samples required manual transfers of the wells of interest from a 1536-well plate into a 384-well plate. Because this manual transfer introduced a source of potential error, it became clear that amore appropriate solutionwould be to sample directly from the 1536-well plates. Currently, commercially available 1536-well plate autosamplers are not compatible with Waters LC/MS systems. The authors have modified their CTC PAL autosampler to support injection from up to twenty-four 1536-well plates. This allows them to cherry-pick any sample from up to 36,864 wells on the autosampler. Because of its success at this Institute, sampling from 1536-well plates has not only become the preferredmethod for LC/MS analysis from IC50 plates but also become the standard format used for the handling of and the sampling from large combinatorial libraries.