Randomness in DNA Encoded Library Selection Data Can Be Modeled for More Reliable Enrichment Calculation

Area of Science:

• Biochemistry
• Chemical Biology
• Drug Discovery

Background:

• DNA Encoded Libraries (DELs) utilize DNA sequences to tag chemical compounds for high-throughput screening.
• Next-generation sequencing allows for the analysis of billions of compounds in DEL libraries.
• Challenges exist in ranking binding events from complex DEL selections due to inherent noise.

Purpose of the Study:

• To investigate the robustness of DNA Encoded Library selection processes.
• To analyze the sources and characteristics of noise in DEL sequencing readouts.
• To develop a framework for improved DEL data analysis.

Main Methods:

• Performed multiple replicate DEL selections against a target protein.
• Examined sequencing readouts of warheads and chemotype families.
• Modeled experimental noise using statistical distributions, specifically the Poisson distribution.

Main Results:

• DEL selection output exhibits intrinsic noise that can be accurately modeled by the Poisson distribution.
• Poisson noise is the primary noise source at low sequence copy counts.
• Noise estimation is feasible even from a single experimental replicate.

Conclusions:

• Directly using copy counts for DEL data analysis can be misleading due to inherent noise.
• A proposed framework incorporating normalization and confidence intervals enhances DEL data interpretation.
• Understanding and modeling Poisson noise is crucial for reliable DEL selection analysis.