Scholarlink

Background <p>Remote homology detection is critical for inferring evolutionary and functional relationships among proteins, but divergence below 30% identity often hinders standard sequence-based searches. Profile-based methods and HMMs improve sensitivity, yet many distant homologues remain undetected. Cascade PSI-BLAST uses efficient, iterative, multi-generation searches of intermediate hits to bridge sequence gaps, constructing successive PSSMs that reveal remote relationships without structural information. However, its application to ultra-large databases such as NR and UniProt is computationally intensive. We address this limitation by optimizing intermediate selection and enabling distributed execution across multiple CPUs or servers, significantly enhancing throughput and resource utilization while preserving detection sensitivity.</p> Results <p>We applied the parallelized Cascade PSI-BLAST algorithm to sequences drawn from multiple SCOP classes to evaluate its sensitivity and predictive power on the GenDis database. Our method uncovers substantially more remote homologues, with less false positives, in both GenDis and UniProt compared to conventional searches. Large repositories such as NR can also be processed in a practical timeframe by distributing cascade searches across multiple servers. Finally, we have deployed a user-friendly web server for running cascade searches on smaller databases, including PDB and Swiss-Prot.</p> Conclusion <p>Our enhanced Cascade PSI-BLAST markedly improves detection of remote homologues across both large and curated databases by leveraging intermediate sequences and distributed execution. Multi-server parallelization reduces runtimes on expansive repositories such as NR to practical levels. The web server offers rapid, user-friendly searches on smaller datasets, while the standalone package supports scalable, customizable analyses on local infrastructure. Together, these tools provide a versatile, sequence-only platform for uncovering distant protein relationships, accelerating functional annotation and evolutionary insights.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cascade PSI-BLAST 2.0: a fast-searching parallelized remote homology detection tool and development of Cascade web server 2.0

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • Soumya Nayak,
  • Dheemanth Reddy Regati,
  • Murugavel Pavalam,
  • R. Sowdhamini

摘要

Background

Remote homology detection is critical for inferring evolutionary and functional relationships among proteins, but divergence below 30% identity often hinders standard sequence-based searches. Profile-based methods and HMMs improve sensitivity, yet many distant homologues remain undetected. Cascade PSI-BLAST uses efficient, iterative, multi-generation searches of intermediate hits to bridge sequence gaps, constructing successive PSSMs that reveal remote relationships without structural information. However, its application to ultra-large databases such as NR and UniProt is computationally intensive. We address this limitation by optimizing intermediate selection and enabling distributed execution across multiple CPUs or servers, significantly enhancing throughput and resource utilization while preserving detection sensitivity.

Results

We applied the parallelized Cascade PSI-BLAST algorithm to sequences drawn from multiple SCOP classes to evaluate its sensitivity and predictive power on the GenDis database. Our method uncovers substantially more remote homologues, with less false positives, in both GenDis and UniProt compared to conventional searches. Large repositories such as NR can also be processed in a practical timeframe by distributing cascade searches across multiple servers. Finally, we have deployed a user-friendly web server for running cascade searches on smaller databases, including PDB and Swiss-Prot.

Conclusion

Our enhanced Cascade PSI-BLAST markedly improves detection of remote homologues across both large and curated databases by leveraging intermediate sequences and distributed execution. Multi-server parallelization reduces runtimes on expansive repositories such as NR to practical levels. The web server offers rapid, user-friendly searches on smaller datasets, while the standalone package supports scalable, customizable analyses on local infrastructure. Together, these tools provide a versatile, sequence-only platform for uncovering distant protein relationships, accelerating functional annotation and evolutionary insights.