Molecular Cues to Smart Sequences: Optimizing Early Round SELEX Sequences

Soniya, Runjhun Saran Narayan, Nived Ambadipudi, Apurva Narayan

Keywords: in-silico, aptamer, SELEX, LLMs

TL;DR: We propose a generative refinement approach that leverages molecular and structural context to guide early-stage aptamer optimization, enabling accelerated discovery through targeted edits that anticipate evolutionary enrichment

Abstract:

Rapid discovery of high-affinity aptamers is often
hindered by low enrichment and high sequence
diversity in early SELEX rounds, leading to prolonged
experimental cycles. To overcome this
challenges, we present a ligand- and structureaware
refinement framework designed to enhance
the quality of aptamer candidates. Our approach
leverages TxGemma, a generative model conditioned
on the ligand’s (small molecule target)
molecular descriptors and the aptamer’s predicted
secondary structure. TxGemma selectively edits
hairpin loops and stems of early-round candidates
to enhance the chance of molecular interactions
with the small molecule ligand. Using
Theophylline as a test case, we show that candidate
sequences from early round (Round 10),
when modified, align closely with the enriched
sequences in later SELEX rounds (Round 16–20).
This demonstrates the model’s ability to emulate
evolutionary convergence and accelerate aptamer
discovery with reduced experimental effort. This
work provides a robust platform for in-depth exploration
and generation of high-affinity aptamers,
potentially eliminating the need for further SELEX
rounds, and accelerate aptamer discovery
with reduced experimental effort.

Submission Number: 161