Abstract
Aptamers are short single-stranded RNA/DNA molecules that bind to specific target molecules. Aptamers with high binding-affinity and target specificity are identified using an in vitro procedure called high throughput systematic evolution of ligands by exponential enrichment (HT-SELEX). However, the development of aptamer affinity reagents takes a considerable amount of time and is costly because HT-SELEX produces a large dataset of candidate sequences, some of which have insufficient binding-affinity. Here, we present RNA aptamer Ranker (RaptRanker), a novel in silico method for identifying high binding-affinity aptamers from HT-SELEX data by scoring and ranking. RaptRanker analyzes HT-SELEX data by evaluating the nucleotide sequence and secondary structure simultaneously, and by ranking according to scores reflecting local structure and sequence frequencies. To evaluate the performance of RaptRanker, we performed two new HT-SELEX experiments, and evaluated binding affinities of a part of sequences that include aptamers with low binding-affinity. In both datasets, the performance of RaptRanker was superior to Frequency, Enrichment and MPBind. We also confirmed that the consideration of secondary structures is effective in HT-SELEX data analysis, and that RaptRanker successfully predicted the essential subsequence motifs in each identified sequence.
Introduction
Aptamers are nucleic acid sequences that fold into complex three-dimensional shapes to bind specific targets with high affinity and specificity. The standard method for discovery, High-Throughput Systematic Evolution of Ligands by Exponential Enrichment (HT-SELEX), involves iterative rounds of selection and amplification to enrich for binders. However, the resulting datasets are massive, often containing millions of sequences.
Traditionally, researchers have relied on sequence frequency in the final round to identify candidates. This approach is flawed because high frequency can result from PCR bias rather than binding affinity. Furthermore, since an aptamer's function is dictated by its shape, ranking methods that ignore secondary structure miss critical information. RaptRanker addresses these limitations by integrating local sequence and structural motifs into a robust ranking framework.
Methods
RaptRanker employs a multi-step computational pipeline to identify high-affinity aptamers from HT-SELEX data, focusing on local structural motifs rather than full-length sequences.
The RaptRanker Pipeline
- Filtering & Pre-processing: Raw sequencing data (FASTA/FASTQ) is cleaned to remove duplicates and ensure sequences conform to the experimental design.
- Sequence-Structure Profiles (SSPs): For each unique sequence, RaptRanker uses the CapR algorithm to calculate a Sequence-Structure Profile. This profile estimates the probability of each nucleotide existing in one of six distinct structural contexts: stem, hairpin loop, bulge, internal loop, multibranch loop, or exterior.
- Subsequence Decomposition: Sequences are broken down into overlapping subsequences using a sliding window. Each subsequence is assigned a Subsequence-Structure Profile (SSSP), focusing the analysis on local binding motifs.
- Clustering via SketchSort: To identify common motifs across millions of sequences, RaptRanker uses SketchSort and a Minimum Spanning Forest to cluster similar SSSPs from all experimental rounds simultaneously.
- Scoring (Average Motif Enrichment): The algorithm calculates an Average Motif Enrichment (AME) score. This metric tracks how much a specific structural motif is enriched across rounds. Sequences containing highly enriched motifs are ranked at the top.
Experimental Validation and Results
The performance of RaptRanker was evaluated using two new HT-SELEX datasets: human recombinant transglutaminase 2 (TG2) and human recombinant integrin alpha V beta 3 (αVβ3). Unlike many studies, these datasets included confirmed low-affinity binders to rigorously test the algorithm's discriminative power.
Superior Ranking Performance
RaptRanker was compared against three established methods: Frequency, Enrichment, and MPBind. In the TG2 dataset, RaptRanker achieved a True Positive Rate (TPR) of 0.524 at a zero False Positive Rate (FPR), vastly outperforming the standard Frequency method (TPR = 0.143). Similar trends were observed in the αVβ3 dataset, even though the binding motifs were shorter and more difficult to identify.
The Importance of Structure
A key finding was that incorporating secondary structure significantly improves identification accuracy. In several instances, structural information alone was a better predictor of binding than sequence alone. Furthermore, RaptRanker successfully predicted the essential binding motifs within sequences, which were later validated through experimental truncation studies.
Conclusion
RaptRanker represents a significant advancement in in silico aptamer selection. By shifting the focus from full-length sequence frequency to the enrichment of local sequence-structure motifs, it provides a more biologically relevant and robust ranking. This approach not only identifies higher-quality candidates but also assists in the optimization of aptamers by pinpointing the minimal active binding core. Future work may involve extending the framework to 3D tertiary structures and developing automated methods to detect and mitigate experimental noise in specific SELEX rounds.
Summary updated