Thomson, M. W. | bioRxiv Science

bioRxiv2026

Spatial second-order features predict glioma malignant transformation

Isocitrate dehydrogenase (IDH) mutant gliomas often transform from low to high grade aggressive tumors. The genetic and molecular drivers of this Malignant Transformation (MT) are poorly understood, and predicting whether a patient will undergo MT is an unmet challenge of high clinical relevance. To stratify patients by MT risk, we applied integrated spatial DNA and RNA profiling to biopsies from 18 retrospective glioma patients which will either remain stable, undergo MT or have already transformed. The resulting dataset consisted of >600,000 single cells, measuring 962 DNA loci, 1150 RNAs, and their spatial locations. Using this dataset, we found that genetic copy number alterations (CNAs), cell types compositions, and cellular neighborhoods did not predict MT. Instead, second order effects i.e. pairwise interactions, are highly predictive of future transformation. First, we identified abnormal chromosomal contact patterns that clearly separate future stable versus future MT samples. Second, we identified 24 ligand-receptors (LR) pairs cross-expressed in neighboring cells as the main molecular factors predictive of transformation and recurrences. We then validated a cross-expressing pair of ENPP2-LPAR1 interactions with a separate cohort of patient samples. In addition, using the LR+ cell pairs as an anchor, we identified cell signaling-specific gene expression programs that can predict from bulk or single cell RNAseq data the time to recurrence. We used a cell-interaction-based foundation model (CIFM) optimized on the spatial RNA data in forward simulations and identified potential myeloid signaling factors involved in MT. Lastly, we analyzed the effect of detection sensitivity on the ability to capture pertinent LR+ neighboring cells by down-sampling transcript and showed that the ability to detect cross-expressing signaling LR transcripts (typically <10 copies per cell) decays rapidly with lower sensitivity, but is more robust to down-sampling of the areas of the tissue imaged. The importance of second-order features suggests that increased depth and dimensionality of data on a smaller quantity of samples can provide valuable insight, and that high-sensitivity and multiple-modalities spatial approaches can help identify markers to risk-stratify patients, aid in therapeutic decision making, and uncover potential therapeutic targets.

cancer biology↗

bioRxiv2023

TRILL: Orchestrating Modular Deep-Learning Workflows for Democratized, Scalable Protein Analysis and Engineering

AO_SCPLOWBSTRACTC_SCPLOWDeep-learning models have been rapidly adopted by many fields, partly due to the deluge of data humanity has amassed. In particular, the petabases of biological sequencing data enable the unsupervised training of protein language models that learn the "language of life." However, due to their prohibitive size and complexity, contemporary deep-learning models are often unwieldy, especially for scientists with limited machine learning backgrounds. TRILL (TRaining and Inference using the Language of Life) is a platform for creative protein design and discovery. Leveraging several state-of-the-art models such as ESM-2, DiffDock, and RFDiffusion, TRILL allows researchers to generate novel proteins, predict 3-D structures, extract high-dimensional representations of proteins, functionally classify proteins and more. What sets TRILL apart is its ability to enable complex pipelines by chaining together models and effectively merging the capabilities of different models to achieve a sum greater than its individual parts. Whether using Google Colab with one GPU or a supercomputer with hundreds, TRILL allows scientists to effectively utilize models with millions to billions of parameters by using optimized training strategies such as ZeRO-Offload and distributed data parallel. Therefore, TRILL not only bridges the gap between complex deep-learning models and their practical application in the field of biology, but also simplifies the orchestration of these models into comprehensive workflows, democratizing access to powerful methods. Documentation: https://trill.readthedocs.io/en/latest/home.html.

bioengineering↗