Biology subjects

Loy, C.

Publications and source records attributed to Loy, C..

3 recordsLinked to original sources

bioRxiv2025

Benchmarking large language models for cell-free RNA diagnostic biomarker discovery

Large-language models (LLMs) can parse vast amounts of data and generate executable code, positioning them as promising tools for the development of biomarkers and classifiers from high-throughput omics data. Here, we benchmarked six LLMs, OpenAIs o3 and GPT-4o, Anthropics Claude Opus 4 and Claude 3.7 Sonnet, and Googles Gemini 2.5 Pro and Gemini 2.0 Flash, for disease classification based on plasma cell-free RNA (cfRNA) profiles obtained by RNA sequencing. We analyzed data from cohorts of children with Kawasaki disease (KD) or multisystem inflammatory syndrome in children (MIS-C), adults with active tuberculosis (TB) or other non-TB respiratory conditions, and individuals with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) or sedentary lifestyle. We assessed two tasks: (i) gene-panel design, where each LLM mined public knowledge to nominate diagnostic genes for use in machine learning (ML), and (ii) end-to-end modeling, where LLMs built an ML workflow directly from raw RNA-seq counts. In the first task, the LLM-derived panels captured canonical immune pathways and outperformed randomly selected genes in all cohorts. They underperformed panels chosen by differential gene expression (DGE) analysis in the KD vs. MIS-C and ME/CFS cohorts but performed comparably or better for the TB cohort. In the second task, o3 produced classifiers for KD vs. MIS-C that performed just as well as conventional statistical methods without human intervention. Performance for TB and ME/CFS cohorts was slightly lower than the conventional approach. These findings delineate current capabilities and limitations of LLMs in diagnostics and open a path for their future use in biomarker discovery.

bioinformatics↗

bioRxiv2024

ByeTAC: Bypassing an E3 Ligase for Targeted Protein Degradation

Targeted protein degradation utilizing a bifunctional molecule to initiate ubiquitination and subsequent degradation by the 26S proteasome has been shown to be a powerful therapeutic intervention. Many bifunctional molecules, including covalent and non-covalent ligands to proteins of interest, have been developed. The traditional target protein degradation methodology targets the protein of interest in both healthy and diseased cell populations, and a therapeutic window is obtained based on the overexpression of the targeted protein. We report here a series of bifunctional degraders that do not rely on interacting with an E3 ligase, but rather a 26S proteasome subunit, which we have named ByeTACs: Bypassing E3 Targeting Chimeras. Rpn-13 is a non-essential ubiquitin receptor for the 26S proteasome. Cells under significant stress or require significant ubiquitin-dependent degradation of proteins for survival, incorporate Rpn-13 in the 26S to increase protein degradation rates. The targeted protein degraders reported here are bifunctional molecules that include a ligand to Rpn-13 and BRD4, the protein of interest we wish to degrade. We synthesized a suite of degraders with varying PEG chain lengths and showed that bifunctional molecules that incorporate a Rpn-13 binder (TCL1) and a BRD4 binder (JQ1) with a PEG linker of 3 or 4 units are the most effective to induce BRD4 degradation. We also demonstrate that our new targeted protein degraders are dependent upon proteasome activity and Rpn-13 expression levels. This establishes a new mechanism of action for our ByeTACs that can be employed for the targeted degradation of a wide variety of protein substrates.

biochemistry↗

bioRxiv2019

Targeted, High-Resolution RNA Sequencing Of Non-Coding Genomic Regions Associated With Neuropsychiatric Functions

The human brain is one of the last frontiers of biomedical research. Genome-wide association studies (GWAS) have succeeded in identifying thousands of haplotype blocks associated with a range of neuropsychiatric traits, including disorders such as schizophrenia, Alzheimers and Parkinsons disease. However, the majority of single nucleotide polymorphisms (SNPs) that mark these haplotype blocks fall within non-coding regions of the genome, hindering their functional validation. While some of these GWAS loci may contain cis-acting regulatory DNA elements such as enhancers, we hypothesized that many are also transcribed into non-coding RNAs that are missing from publicly available transcriptome annotations. Here, we use targeted RNA capture ( RNA CaptureSeq) in combination with nanopore long-read cDNA sequencing to transcriptionally profile 1,023 haplotype blocks across the genome containing non-coding GWAS SNPs associated with neuropsychiatric traits, using post-mortem human brain tissue from three neurologically healthy donors. We find that the majority (62%) of targeted haplotype blocks, including 13% of intergenic blocks, are transcribed into novel, multi-exonic RNAs, most of which are not yet recorded in GENCODE annotations. We validated our findings with short-read RNA-seq, providing orthogonal confirmation of novel splice junctions and enabling a quantitative assessment of the long-read assemblies. Many novel transcripts are supported by independent evidence of transcription including cap analysis of gene expression (CAGE) data and epigenetic marks, and some show signs of potential functional roles. We present these transcriptomes as a preliminary atlas of non-coding transcription in human brain that can be used to connect neurological phenotypes with gene expression.

genomics↗