Improvements in viral gene annotation using large language models and soft alignments

BACKGROUND

The annotation of protein sequences in public databases has long posed a challenge in molecular biology. This issue is particularly acute for viral proteins, which demonstrate limited homology to known proteins when using alignment, k-mer, or profile-based homology search approaches. A novel methodology employing Large Language Models (LLMs) addresses this methodological challenge by annotating protein sequences based on embeddings.

RESULTS

Central to our contribution is the soft alignment algorithm, drawing from traditional protein alignment but leveraging embedding similarity at the amino acid level to bypass the need for conventional scoring matrices. This method not only surpasses pooled embedding-based models in efficiency but also in interpretability, enabling users to easily trace homologous amino acids and delve deeper into the alignments. Far from being a black box, our approach provides transparent, BLAST-like alignment visualizations, combining traditional biological research with AI advancements to elevate protein annotation through embedding-based analysis while ensuring interpretability. Tests using the Virus Orthologous Groups and ViralZone protein databases indicated that the novel soft alignment approach recognized and annotated sequences that both blastp and pooling-based methods, which are commonly used for sequence annotation, failed to detect.

CONCLUSION

The embeddings approach shows the great potential of LLMs for enhancing protein sequence annotation, especially in viral genomics. These findings present a promising avenue for more efficient and accurate protein function inference in molecular biology.

Analyzing Glycan-Binding Profiles Using Weighted Multiple Alignment of Trees

This chapter describes the Multiple Carbohydrate Alignment with Weights (MCAW) tool, which is available as a part of the RINGS (Resource for INformatics of Glycomes at Soka) website. It implements a combination of KCaM (Aoki, Yamaguchi, Ueda, et al., Nucl Acids Res 32:W267-W272, 2004), a pairwise glycan alignment algorithm, and ClustalW (Thompson, Higgins, Gibson, Nucleic Acids Res 22:4673-80, 1994), a weighted multiple protein sequence alignment algorithm. This tool computes the multiple glycan alignment by first computing a guide tree to determine the order by which to progressively add glycans to the multiple alignment. The dynamic programming algorithm results in a glycan profile of the alignment glycans, containing "monosaccharide positions" indicating the ratio of monosaccharides and their glycosidic bonds that are aligned at the corresponding position. This tool has been used to analyze databases of glycan array experimental data, incorporating weights to reflect the biological significance of certain glycans over others. As a result, it has been shown that the alignments obtained are biologically relevant, matching the results as found in the literature.

Towards a Consistent, Quantitative Evaluation of MicroRNA Evolution

The miRBase currently reports more than 25,000 microRNAs in several hundred genomes that belong to more than 1000 families of homologous sequences. Quantitative investigations of miRNA gene evolution requires the construction of data sets that are consistent in their coverage and include those genomes that are of interest in a given study. Given the size and structure of data, this can be achieved only with the help of a fully automatic pipeline that improves the available seed alignments, extends the set of available sequences by homology search, and reliably identifies true positive homology search results. Here we describe the current progress towards such a system, emphasizing the task of improving and completing the initial seed alignment.

Evaluation and assessment of read-mapping by multiple next-generation sequencing aligners based on genome-wide characteristics

Massive data produced due to the advent of next-generation sequencing (NGS) technology is widely used for biological researches and medical diagnosis. The crucial step in NGS analysis is read alignment or mapping which is computationally intensive and complex. The mapping bias tends to affect the downstream analysis, including detection of polymorphisms. In order to provide guidelines to the biologist for suitable selection of aligners; we have evaluated and benchmarked 5 different aligners (BWA, Bowtie2, NovoAlign, Smalt and Stampy) and their mapping bias based on characteristics of 5 microbial genomes. Two million simulated read pairs of various sizes (36bp, 50bp, 72bp, 100bp, 125bp, 150bp, 200bp, 250bp and 300bp) were aligned. Specific alignment features such as sensitivity of mapping, percentage of properly paired reads, alignment time and effect of tandem repeats on incorrectly mapped reads were evaluated. BWA showed faster alignment followed by Bowtie2 and Smalt. NovoAlign and Stampy were comparatively slower. Most of the aligners showed high sensitivity towards long reads (>100bp) mapping. On the other hand NovoAlign showed higher sensitivity towards both short reads (36bp, 50bp, 72bp) and long reads (>100bp) mappings; It also showed higher sensitivity towards mapping a complex genome like Plasmodium falciparum. The percentage of properly paired reads aligned by NovoAlign, BWA and Stampy were markedly higher. None of the aligners outperforms the others in the benchmark, however the aligners perform differently with genome characteristics. We expect that the results from this study will be useful for the end user to choose aligner, thus enhance the accuracy of read mapping.

Data on four apoptosis-related genes in the colonial tunicate Botryllus schlosseri

The data described are related to the article entitled “Recurrent phagocytosis-induced apoptosis in the cyclical generation change of the compound ascidian Botryllus schlosseri” (Franchi et al., 2016) [1]. Four apoptosis-related genes, showing high similarity with mammalian Bax (a member of the Bcl-2 protein family), AIF1 (apoptosis-inducing factor-1), PARP1 (poly ADP ribose polymerase-1) and IAP7 (inhibitor of apoptosis-7) were identified from the analysis of the trascriptome of B. schlosseri. They were named BsBax, BsAIF1, BsPARP1 and BsIAP7. Here, their deduced amino acid sequence were compared with known sequences of orthologous genes from other deuterostome species together with a study of their identity/similarity.