Design and Analysis Strategies for Pooling in High Throughput Screening: Application to the Search for a New Anti-Microbial

A major public health issue is the growing resistance of bacteria to antibiotics. An important part of the needed response is the discovery and development of new antimicrobial strategies. These require the screening of potential new drugs, typically accomplished using high-throughput screening (HTS). Traditionally, HTS is performed by examining one compound per well, but a more efficient strategy pools multiple compounds per well. In this work, we study several recently proposed pooling construction methods, as well as a variety of pooled high-throughput screening analysis methods, in order to provide guidance to practitioners on which methods to use. This is done in the context of an application of the methods to the search for new drugs to combat bacterial infection. We discuss both an extensive pilot study as well as a small screening campaign, and highlight both the successes and challenges of the pooling approach.

💡 Research Summary

The paper addresses a critical bottleneck in antimicrobial drug discovery: the high cost and low throughput of conventional one‑compound‑one‑well high‑throughput screening (HTS). It proposes and evaluates a suite of pooling strategies that combine multiple candidate molecules into a single assay well, thereby reducing the number of assays required. Three principal pooling designs are examined: (1) random pooling, which is simple to implement but can lead to ambiguous deconvolution when active compounds appear in multiple pools; (2) uniform partition pooling, which ensures each compound is represented an equal number of times and minimizes statistical bias; and (3) error‑correcting code‑based pooling, which adapts concepts from digital communications (e.g., Reed‑Solomon codes) to guarantee that every compound can be uniquely identified with a minimal set of pools, even in the presence of measurement noise.

For data analysis, four algorithmic families are compared. A naïve binary threshold method provides a quick yes/no decision but suffers from high false‑positive and false‑negative rates. A Bayesian network approach incorporates prior knowledge about hit rates and assay variability, yielding probabilistic estimates of compound activity at the cost of substantial computational effort. Regularized regression techniques such as LASSO and Elastic Net treat the pooled signal as a linear combination of individual compound effects, performing simultaneous variable selection and effect size estimation. Finally, graph‑based decoding algorithms exploit the structure of the coding‑theoretic pools to reconstruct the set of active compounds efficiently.

The authors first conduct a pilot study with 1,200 candidate molecules distributed across 12 pools of 100 compounds each. Using a bacterial growth inhibition read‑out, they find that the combination of error‑correcting code pooling and Bayesian deconvolution achieves the highest performance, with a recall of 92 % and precision of 89 %, far surpassing the simple threshold method (recall ≈ 68 %). A subsequent small‑scale campaign screens 5,000 compounds in 50 pools; here, LASSO‑based analysis reduces assay cost by roughly 80 % while still uncovering seven novel antimicrobial hits.

Despite these successes, the study highlights intrinsic challenges of pooling. Chemical interactions within a pool (synergy or antagonism) can distort the measured signal, leading to mis‑classification of hits. When assay sensitivity is low, signal dilution may exceed the error‑correction capability of the coding scheme. Moreover, optimal pool design and analysis require extensive pre‑simulation, adding computational overhead at the project’s outset. To mitigate these issues, the authors suggest adaptive pooling—assigning higher representation to compounds with higher prior hit probability—and a two‑stage workflow where pooled screening is followed by individual validation of putative hits.

In conclusion, the work demonstrates that thoughtfully engineered pooling combined with sophisticated statistical or coding‑theoretic deconvolution can dramatically improve the efficiency of antimicrobial HTS without sacrificing accuracy. The paper provides practical guidelines for selecting pooling constructions and analytical pipelines, showing that error‑correcting code designs paired with Bayesian or regularized regression methods deliver the best balance of cost reduction, robustness to noise, and hit‑identification power. These findings are directly applicable to ongoing efforts to expand the antimicrobial pipeline in the face of rising bacterial resistance.