Biology Meets Computation: Papers Bridging Life and Data Science

Yue, Dai, Tang, Zhou, Mathews & Huang (2026). mRNA vaccine design requires simultaneously optimising sequence for codon usage, secondary structure stability and immunogenicity. This paper casts mRNA design as a continuous optimisation problem and applies differentiable sampling methods that navigate the complex multi-objective landscape more efficiently than prior combinatorial approaches.

Liu, Yuan, Rao, Reddy & Jacobs (2026). Intrinsically disordered proteins (IDPs) drive liquid-liquid phase separation in cells — the process underlying stress granule formation, chromatin organisation, and increasingly, neurodegenerative disease. This paper introduces a thermodynamic metric that predicts how IDPs partition into condensates, providing a quantitative design rule for synthetic biology.

Catrina, Bepler, Sledzieski & Singh (2026). Protein language models trained on evolutionary databases encode rich structural and functional information, but their representations become inconsistent as model scale increases. Reverse distillation — using larger models to supervise smaller ones in a direction that preserves biological consistency — provides a principled solution to scaling without quality degradation.

Gu, Gao, He, Zhang, Wen, Luo, Wang, Cao, Bu & Hsieh (2026). Chemical modifications to proteins, RNA and DNA are biologically critical but are systematically under-represented in training data for biomolecular models. Bi-TEAM introduces cross-scale contrastive learning that bridges atomic-level modification representations with sequence-level functional annotations.

Swami, McBride, Eckmann & Tlusty (2026). Comparing protein surfaces requires a distance metric that captures both geometric shape and chemical character simultaneously. This paper defines a joint metric using Riemannian geometry on the protein surface manifold, enabling more accurate pocket-to-pocket comparisons for drug binding site analysis than either geometry or chemistry alone.

Vilar, Rubi & Saiz (2026). Derives scaling laws for the entropic separation of biopolymers (DNA, RNA, proteins) through nanoscale filters — a process critical for gel electrophoresis and nanopore sequencing. The paradoxical metastable states observed at certain scale regimes explain previously unexplained anomalies in experimental separation profiles.

Geddes-Nelson, Liu & Yong (2026). Huntington's disease is caused by polyglutamine tract expansion in the huntingtin protein, but the conformational mechanism linking tract length to aggregation has been poorly understood. This molecular dynamics study reveals length-dependent structural transitions and identifies co-solvent conditions that modulate aggregation propensity — potential leads for therapeutic intervention.

Zhao, Zhang, Guo & Li (2026). A methodological critique showing that conventional ancient DNA analysis methods systematically violate preservation constraints that limit which molecular signals can be reliably recovered from degraded samples. The proposed HSF framework applies Bayesian posterior sourcing to distinguish genuine ancient genetic information from contamination and decay artefacts.

Gu, Gao, He, Zhang, Wen, Luo, Wang, Cao, Bu & Hsieh (2026). Chemical modifications such as phosphorylation, methylation, and glycosylation alter biomolecule function in ways that sequence alone cannot capture. Bi-TEAM's cross-scale contrastive learning bridges atomic-resolution modification representations with sequence-level functional annotations, enabling prediction of modification effects without requiring experimentally labelled modification sites.

Yue, Dai, Tang, Zhou, Mathews & Huang (2026). mRNA vaccine design requires simultaneously optimising sequence for codon usage, secondary structure stability and immunogenicity. This paper casts mRNA design as a continuous optimisation problem and applies differentiable sampling methods that navigate the complex multi-objective landscape more efficiently than prior combinatorial approaches.

Liu, Yuan, Rao, Reddy & Jacobs (2026). Intrinsically disordered proteins (IDPs) drive liquid-liquid phase separation in cells — the process underlying stress granule formation, chromatin organisation, and increasingly, neurodegenerative disease. This paper introduces a thermodynamic metric that predicts how IDPs partition into condensates, providing a quantitative design rule for synthetic biology.

Catrina, Bepler, Sledzieski & Singh (2026). Protein language models trained on evolutionary databases encode rich structural and functional information, but their representations become inconsistent as model scale increases. Reverse distillation — using larger models to supervise smaller ones in a direction that preserves biological consistency — provides a principled solution to scaling without quality degradation.

Gu, Gao, He, Zhang, Wen, Luo, Wang, Cao, Bu & Hsieh (2026). Chemical modifications to proteins, RNA and DNA are biologically critical but are systematically under-represented in training data for biomolecular models. Bi-TEAM introduces cross-scale contrastive learning that bridges atomic-level modification representations with sequence-level functional annotations.

Swami, McBride, Eckmann & Tlusty (2026). Comparing protein surfaces requires a distance metric that captures both geometric shape and chemical character simultaneously. This paper defines a joint metric using Riemannian geometry on the protein surface manifold, enabling more accurate pocket-to-pocket comparisons for drug binding site analysis than either geometry or chemistry alone.

Vilar, Rubi & Saiz (2026). Derives scaling laws for the entropic separation of biopolymers (DNA, RNA, proteins) through nanoscale filters — a process critical for gel electrophoresis and nanopore sequencing. The paradoxical metastable states observed at certain scale regimes explain previously unexplained anomalies in experimental separation profiles.

Geddes-Nelson, Liu & Yong (2026). Huntington's disease is caused by polyglutamine tract expansion in the huntingtin protein, but the conformational mechanism linking tract length to aggregation has been poorly understood. This molecular dynamics study reveals length-dependent structural transitions and identifies co-solvent conditions that modulate aggregation propensity — potential leads for therapeutic intervention.

Zhao, Zhang, Guo & Li (2026). A methodological critique showing that conventional ancient DNA analysis methods systematically violate preservation constraints that limit which molecular signals can be reliably recovered from degraded samples. The proposed HSF framework applies Bayesian posterior sourcing to distinguish genuine ancient genetic information from contamination and decay artefacts.

Gu, Gao, He, Zhang, Wen, Luo, Wang, Cao, Bu & Hsieh (2026). Chemical modifications such as phosphorylation, methylation, and glycosylation alter biomolecule function in ways that sequence alone cannot capture. Bi-TEAM's cross-scale contrastive learning bridges atomic-resolution modification representations with sequence-level functional annotations, enabling prediction of modification effects without requiring experimentally labelled modification sites.