Sign in Subscribe
Machine learning

State-of-the-Art Machine Learning Tools: Transforming Genetic Data Analysis

Strategic Insights

— 1 min read

Researchers at Calico Life Sciences LLC have made significant breakthroughs in using state-of-the-art machine learning tools to predict gene expression data from DNA sequences. By leveraging ideas from signal-processing literature, the team has developed new approaches inspired by state-space models, Fourier transforms, and wavelet transforms, which have shown promising results in functional genomics.

Key Takeaways:

  • The team has developed a software library called bilby, implemented using Python and Jax/Flax, which provides a range of architectures for supervised multi-task learning in functional genomics, including convolutional, attention, bidirectional Hyena, bidirectional Mamba, and striped-architecture models.
  • The study compared the performance of these architectures, testing several hyperparameters and variations, and reported performance statistics for the withheld test set as well as downstream SNP classifiers.
  • Models comprising convolutional, bidirectional Mamba, and (optionally) attention layers achieved small but consistent improvements in prediction accuracy, for roughly comparable training times and parameter counts, when averaged across all output tracks and data splits.
  • The gains in prediction accuracy of the Mamba-based models do not yet translate to significantly improved performance on downstream SNP classification tasks.
  • The research indicates that selective state-space models, such as Mamba and Striped Mamba, warrant further exploration for functional genomics tasks.

Statistics:

  • The Mamba-based models achieved a proportional increase of 3-4% in Pearson R and 1-2% in r, with the highest gains achieved when Mamba and attention layers were combined in a striped architecture.
  • The study reported a difference of +0.007 in area under ROC and -0.006 in Spearman rank correlation between Mamba- and attention-based classifiers on the GTEx eQTL dataset.
  • The research has been peer-reviewed and publicly available at https://github/com/calicolife/bilby.

Sources:

  • NewsRx. Findings in the Area of Computers Reported from Calico Life Sciences LLC (Selective State Space Models Outperform Transformers at Predicting RNA-Seq Read Coverage). Computer Weekly News. October 29, 2025; p 187.
  • Johannes Linder, Calico Life Sciences LLC, 1170 Veterans Blvd, South San Francisco, CA 94080.
  • Selective State Space Models Outperform Transformers at Predicting RNA-Seq Read Coverage. bioRxiv, 2025.

Read more