Amplifying DNA or RNA is a key step in many areas of biology, from disease diagnostics to emerging technologies like DNA-based data storage. However, controlling how efficiently genetic material is copied during amplification is not always easy. Small changes in sequence or reaction conditions can lead to uneven results, which can limit sensitivity and accuracy.

Researchers from Shanghai Jiao Tong University introduce a new way to precisely control amplification efficiency using a thermodynamics-based design strategy. The team combined experimental biology with computational modeling to address this long-standing challenge.

Overview of Programmable Amplification

(A) In conventional multiplex amplification systems, uniform amplification often leads to loss of low-frequency templates. In contrast, PA enables template-specific amplification tuning, facilitating both enrichment of low-abundance templates and on-demand multitarget analysis. (B) Workflow of PA experiment, comprising two key steps: signal recognition (nucleic acid hybridization) and enrichment step (enzymatic amplification). (C) Relationship between equivalent amplification efficiency (abbreviated Equivalent ampl. eff.) and primer hybridization thermodynamics in conventional methods. The narrow ∆G tuning range results in abrupt efficiency changes upon single-base modifications. (D) Thermodynamic regulation in PA shows an expanded ∆G tuning range, allowing precise programmability of amplification efficiency. (E) Schematic of the machine learning–assisted efficiency prediction model for PA.

The core idea is to separate two factors that are usually linked together, sequence recognition and binding strength. By using a primer-tag compensation strategy, the researchers were able to fine-tune how strongly DNA strands bind during amplification without changing which sequences are targeted. This allows amplification to be programmed with much higher precision than traditional approaches.

To further improve performance, the team built a machine learning model using more than 2,400 experimental data points. This model significantly improved the ability to predict amplification outcomes, making the system more reliable and easier to design. The improvement in predictive accuracy was substantial, showing how combining biology with data science can enhance experimental control.

The impact of this method is broad. In DNA data storage, it enables higher information density and introduces new ways to hide data using controlled amplification differences. In clinical applications, the method improved detection of rare RNA variants linked to cervical cancer. It was able to detect signals that were previously missed, increasing sensitivity by up to 100-fold under simulated sequencing conditions.

This approach also has potential for advanced applications such as RNA sequencing, single-cell analysis, and spatial transcriptomics, where precise control over amplification can improve data quality and resolution. Overall, the work provides a new framework for making nucleic acid amplification more predictable, tunable, and powerful across a range of biological and technological fields.

Availability – The code used in this study is available on GitHub (https://github.com/NABMElab/programmable-amplification)