Ghodsi, A. | bioRxiv Science

bioRxiv2026

Deep learning-based non-invasive profiling of tumor transcriptomes from cell-free DNA for precision oncology

Tumor gene expression profiling provides crucial diagnostic information for guiding therapy, but standard tissue biopsies are invasive, spatially biased, and may inadequately sample metastatic disease. Cell-free DNA (cfDNA) provides a minimally invasive alternative for tumor genotyping, yet reconstructing robust, transcriptome-wide expression from standard-depth cfDNA whole-genome sequencing (WGS) remains a major challenge. We developed a deep learning framework comprising Triton, for comprehensive cfDNA feature extraction, and Proteus, a probabilistic model that infers single-gene expression from standard-depth cfDNA WGS. Proteus outperformed prior cfDNA approaches in reconstructing molecular phenotypes from matched tumor transcriptomes across multiple cancer types, including prostate, lung, and bladder cancer cohorts, with uncertainty-guided withholding improving model reliability. Proteus further enabled assessment of therapeutic target activity, prognostic transcriptional programs, and candidate treatment-emergent resistance states, establishing a generalizable framework for minimally invasive functional genomics in precision oncology.

bioinformatics↗

bioRxiv2024

Conditional Sequence-Structure Integration: A Novel Approach for Precision Antibody Engineering and Affinity Optimization

Antibodies, or immunoglobulins, are integral to the immune response, playing a crucial role in recognizing and neutralizing external threats such as pathogens. However, the design of these molecules is complex due to the limited availability of paired structural antibody-antigen data and the intricacies of structurally non-deterministic regions. In this paper, we introduce a novel approach to designing antibodies by integrating structural and sequence information of antigens. Our approach employs a protein structural encoder to capture both sequence and conformational details of antigen. The encoded antigen information is then fed into an antibody language model (aLM) to generate antibody sequences. By adding cross-attention layers, aLM effectively incorporates the antigen information from the encoder. For optimal model training, we utilized the Causal Masked Language Modeling (CMLM) objective. Unlike other methods that require additional contextual information, such as epitope residues or a docked antibody framework, our model excels at predicting the antibody sequence without the need for any supplementary data. Our enhanced methodology demonstrates superior performance when compared to existing models in the RAbD benchmark for antibody design and SKEPMI for antibody optimization.

immunology↗