Aptamers are short synthetic nucleic acids that form complex three-dimensional structures, enabling them to bind to target molecules. The binding properties of aptamers have attracted considerable interest in using them for molecular perturbations, detection of disease-related targets, and as tools in synthetic biology. Aptamers are identified through a process known as systematic evolution of ligands by exponential enrichment (SELEX) (1, 2). SELEX is an iterative selection method that uses large libraries of nucleic acids. During this process, the aptamer sequences within these libraries that bind to target molecules are enriched. Nevertheless, it remains a challenging task to identify the most effective, target-specific aptamers among the enriched sequences. On page 51 of this issue, Luo et al. describe single-cell perturbation-driven aptamer recognition and kinetics sequencing (SPARK-seq), a method that combines the binding properties and sequencing capabilities of aptamers with gene inactivation studies (3). This approach enables the simultaneous mapping of thousands of aptamer-target interactions and the identification of aptamers that bind to low-abundance targets.

Because aptamers are nucleic acids, they can be easily detected and quantified by polymerase chain reaction (PCR). This distinguishes aptamers from other molecular binders, for example, antibodies or small molecules, and facilitates the quantification of aptamers and their bound targets. SELEX involves repetitive cycles of incubation of nucleic acid libraries with a target, the separation of bound from unbound sequences, and the amplification of the bound sequences by PCR. Using living cells as target structures enables the identification of aptamers that bind to receptors, other transmembrane proteins, and proteins located on the extracellular surface of the cell (4, 5). Aptamers identified in this manner are typically used for cell-specific administration of (chemo)therapeutics or for target identification. For the latter, proteomic methods are used to determine the specific protein to which an aptamer binds. However, this approach is limited to small sample numbers owing to the time-consuming nature of the process and the necessity to optimize the procedure for each aptamer-target pair. Thus, target identification focuses on a handpicked group of aptamers that are often the most abundant in enriched libraries and tend to bind to highly expressed proteins.

Aptamers that bind to low-abundance targets are rarely found in enriched libraries (6). In addition to binding properties, the intrinsic characteristics of nucleic acids also affect the frequency of aptamers in these libraries. For example, the efficiency with which nucleic acids are amplified by PCR depends on the composition and structural intricacies of a particular sequence. This efficiency can influence the copy number of enriched sequences, independent of their binding properties. The advent of deep-sequencing methods has made extensive sequence information from enriched libraries of SELEX experiments available (7). This development has opened the door for the advancement of bioinformatic tools to assess these data and for artificial intelligence approaches to improve the binding properties and specificity of aptamers (8). These advancements can help to identify the best available aptamers in the enriched libraries. Nevertheless, it is evident that a substantial gap exists between the latent potential inherent in enriched libraries and the ability to harness this potential by identifying and assigning to a target all conceivable aptamers.

SPARK-seq addresses some of the challenges of identifying and assigning aptamers by the integration of aptamer features with genetic perturbations. Luo et al. generated an aptamer library that binds to human breast cancer cells (SUM159). Subsequently, the aptamer library was incubated with cells in which individual genes had been inactivated by CRISPR technologies. Single-cell sequencing was used to provide information on the cell-bound aptamer population and the CRISPR guide RNAs present . Bioinformatic analyses were then used to correlate the loss-of-aptamer binding phenotypes with the CRISPR-based gene inactivation of a target protein (see the figure) . Clustering and statistical methods were applied to sequence data to identify families of aptamers that bind to specific target proteins. SPARK-seq was able to identify aptamer targets that were minimally abundant, and Luo et al. validated this target binding in vitro. Notably, these aptamers and targets are usually challenging to detect but are of particular interest for diagnostic and therapeutic applications.

Luo et al. not only mapped more than 5000 aptamers to eight different target proteins but also reported information on numerous nonbinding sequences contained in the libraries. This information is of substantial value for the use of large language models (a type of artificial intelligence) to design aptamer sequences with improved kinetic properties, as demonstrated by the single-cell perturbationaptamer recognition and targeted aptamer-generation algorithm (SPARTA) that was developed and used in the study.

Luo et al. applied SPARK-seq to target proteins that were known to be localized on the cell surface. However, future studies might be able to apply the method to known proteins that have not yet been associated with a cell-surface presence. Furthermore, Luo et al. focused SPARK-seq on the 10,000 most prevalent sequences in the enriched library, which cover more than 60% of the total sequence population. This sequence distribution is consistent with other studies but suggests that the library contains additional aptamers that bind to other targets. Owing to their low frequency, identifying these sequences remains a statistical challenge and would necessitate refined SELEX methods that suppress the enrichment of aptamers that bind to common targets.

Looking forward, the approach reported by Luo et al. could be applied in various ways, including adaptation to examine different cellular states, such as cancer differentiation, or differences in cell surface proteomes that result from pharmacological perturbations. Consequently, integrating this approach with functional measurements could enable the assignment of aptamers to targets and the modulation of biological functions, including the inhibition or activation of signaling cascades, that extend beyond the scope of pure binding. Similarly, the incorporation of chemically modified libraries or noncanonical base pairs into the SPARK-seq procedure could substantially expand the number of addressable target proteins (9–13).

Aptamer libraries enriched by SELEX experiments with tissue or plasma samples have been used for the classification of cancer patient cohorts (6, 14, 15). However, these libraries either bind to presently unknown or relatively abundant targets. Thus, channeling these libraries into the SPARK-seq workflow could advance the identification of targets and aptamers that are associated with the respective cancer classification—information that could be useful for precision medicine approaches.

Luo et al. introduce a promising strategy that combines aptamer technology with data science, rendering aptamers suitable for systems biology approaches. By examining a multitude of aptamer-target pairs, SPARK-seq advances aptamer identification and overcomes some prevailing limitations. The authors provide an alternative perspective on aptamer technology, establishing a foundation for using aptamer sequence information in omics-like workflows. This enables the integration of aptamers into comprehensive, data-driven technologies, leveraging their inherent molecular interaction characteristics.