⚙
AI-generated deep summary
by claude@2026-07, 2026-07-15
· read from full text
ⓘ
This paper presents APIPred Web 1.0, a user-facing web platform that takes a target protein sequence and an aptamer template (PREFIX–[VARIABLE]–SUFFIX), then generates constraint-aware aptamer libraries, predicts aptamer–protein interactions with a trained XGBoost model, and computes DNA secondary structure minimum free energy using ViennaRNA. The system converts aptamer candidates into k-mer features and proteins into PAAC descriptors, uses precomputed protein features plus vectorized batch processing, retains only the top 25 candidates during generation, and returns ranked sequences with log-transformed interaction scores, dot-bracket structures, and interactive 2D visualizations. As a demonstration, it generated 25 putative CD64 binders from a custom 40-nt library, where flow cytometry reported specific binding to CD64-expressing THP-1 cells with minimal signal in Ramos controls. The paper does not explicitly discuss endometriosis or adenomyosis; it was included in the corpus via a keyword match in the upstream search index.
Abstract
Aptamers are short single-stranded nucleic acids that bind protein targets with high specificity and are increasingly used in diagnostics and therapeutics, yet experimental discovery remains slow and variable in success. This creates a demand for computational systems that not only score candidate binders but also generate experimentally usable libraries under biologically meaningful constraints. Here, we present APIPred Web 1.0, a unified web platform that integrates constraint-aware aptamer library generation, machine learning- based aptamer- protein interaction prediction, and DNA secondary-structure analysis within a user-facing workflow. Users submit a target protein aminoacid sequence and define an aptamer template in a PREFIX - [VARIABLE] - SUFFIX format with real-time validation of key biological constraints (GC content and homopolymer limits). On the backend, sequences are converted into model- compatible features via optimized k-mer encodings (aptamer) and pseudo amino acid composition (PAAC) descriptors (protein), followed by inference with a trained XGBoost predictor. APIPred Web 1.0 improves the computational efficiency by applying precomputed protein features, vectorized batch processing (hundreds of sequences per batch), optimized XGBoost DMatrix inference, and a bounded heap that retains only the top 25 candidates during generation. The platform then computes minimum free energy (MFE) structures using ViennaRNA with parallel folding and returns ranked list of the top candidates with log-transformed interaction scores, complete sequences (variable region highlighted), dot-bracket structures, MFE values, and interactive 2D visualizations via persistent result links. In a demonstration study targeting CD64 protein, the platform produced 25 putative binders from a custom 40- nucleotide library and enabled selection of structurally diverse candidates for experimental testing. Flow cytometry showed specific binding to CD64-expressing THP-1 cells with minimal signal in Ramos control cells. Collectively, APIPred Web 1.0 offers a reproducible, structure-informed, and computationally efficient pipeline for rapid generation of aptamer candidates against target proteins for downstream experimental validation. Graphical Abstract
Full text
2,339 characters
· extracted from
oa-doi-fallback
· click to expand
Abstract
Aptamers are short single-stranded nucleic acids that bind protein targets with high specificity and are increasingly used in diagnostics and therapeutics, yet experimental discovery remains slow and variable in success. This creates a demand for computational systems that not only score candidate binders but also generate experimentally usable libraries under biologically meaningful constraints. Here, we present APIPred Web 1.0, a unified web platform that integrates constraint-aware aptamer library generation, machine learning- based aptamer- protein interaction prediction, and DNA secondary-structure analysis within a user-facing workflow. Users submit a target protein aminoacid sequence and define an aptamer template in a PREFIX - [VARIABLE] - SUFFIX format with real-time validation of key biological constraints (GC content and homopolymer limits). On the backend, sequences are converted into model- compatible features via optimized k-mer encodings (aptamer) and pseudo amino acid composition (PAAC) descriptors (protein), followed by inference with a trained XGBoost predictor. APIPred Web 1.0 improves the computational efficiency by applying precomputed protein features, vectorized batch processing (hundreds of sequences per batch), optimized XGBoost DMatrix inference, and a bounded heap that retains only the top 25 candidates during generation. The platform then computes minimum free energy (MFE) structures using ViennaRNA with parallel folding and returns ranked list of the top candidates with log-transformed interaction scores, complete sequences (variable region highlighted), dot-bracket structures, MFE values, and interactive 2D visualizations via persistent result links. In a demonstration study targeting CD64 protein, the platform produced 25 putative binders from a custom 40- nucleotide library and enabled selection of structurally diverse candidates for experimental testing. Flow cytometry showed specific binding to CD64-expressing THP-1 cells with minimal signal in Ramos control cells. Collectively, APIPred Web 1.0 offers a reproducible, structure-informed, and computationally efficient pipeline for rapid generation of aptamer candidates against target proteins for downstream experimental validation.
Competing Interest Statement
The authors have declared no competing interest.
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.