Jiangning (John) Song

Director of Data-driven Bioinformatics and Biomedicine Lab | Professor of Monash University | Assoc Editor of IEEE J Biomed Health Infomatics, BMC Bioinform

[Jiangning (John) Song @supercs08]

We have 2 open RA+PhD positions in our AI-driven Bioinformatics & Biomed Lab @MonashBDI, for onshore students with strong CS, SE/ EE background & interest in medical imaging, starting any time soon, supported by Monash Major & Seed IDR Grants. Welcome to contact me if interested.

[César de la Fuente @delafuentelab]

New preprint out! We introduce Termini, our AI framework for peptide design.

Termini is a modification-aware generative framework that designs peptides with explicit N- and C-terminal modifications and predicts activity across 15 bacterial species.

To validate the approach, we synthesized and tested 120 peptides across 11 pathogens. 111/120 (92.5%) showed in vitro antimicrobial activity, and terminal modifications consistently improved potency.

Lead candidates also showed in vivo anti-infective efficacy, highlighting the potential of terminally modified, AI-designed peptide antibiotics.

More broadly, this work points toward a future in which peptide design becomes increasingly programmable.

The era of peptide design is here.

Great team work Jing Xu, Chen Li, Jian Li, @supercs08 @MonashUni, @mdt_torres, Fuyi Li @UniofAdelaide, @Penn @PennBioeng @PennEngineers @PennMedicine @PennChemistry @CBE_Penn @PennMicro @PennPsych @PennSAS

Link to paper: https://t.co/xnyE09XGcx

[César de la Fuente @delafuentelab]

Excited to share our new review in @CellBiomat @CellPressNews on how AI is reshaping antibiotic discovery.

We review predictive and generative approaches for discovering both small-molecule antibiotics and antimicrobial peptides—from screening and optimization to de novo design—highlighting protein language models, GNNs, current challenges, and future opportunities.

Great collaboration with Jing Xu, Chen Li, Xiaoyu Wang, Anton Y. Peleg, @supercs08 @MonashUni, Fuyi Li, @UniofAdelaide, @Penn, @PennBioeng, @PennEngineers, @PennMedicine, @PennChemistry, @CBE_Penn, @PennMicro, @PennPsych, and @PennSAS

https://t.co/ZOnfexpgNI

[Nature Methods @naturemethods]

Also read the N&V highlighting the work here:

https://t.co/aLOFhylkG3

[Nature Methods @naturemethods]

DECODE: a universal deconvolution framework for transcriptomic, proteomic and metabolomic data.

https://t.co/rYkggtwz2G

[César de la Fuente @delafuentelab]

Check out our @NatureBiotech @NaturePortfolio News&Views with colleagues Yumeng Zhang and @supercs08 on @genentech's deep learning approach to accelerate antibiotic discovery. Congrats to Gabriele Scalia, @tbyanc, Man-Wah Tan, Aviv Regev, @Yoshua_Bengio and team on this great progress!

[Nature Biotechnology @NatureBiotech]

Deep learning speeds the search for new antibiotic scaffolds #NBTNV https://t.co/NDyslQ7Pp0

[Nature Computational Science @NatComputSci]

📢Xiangliang Zhang and colleagues evaluate bias in AI-generated medical text, revealing disparities across race, sex, and age, and propose an optimization method to enhance fairness without compromising performance. https://t.co/3vi3F4Id30

🔓https://t.co/XiBWNVnFay

[Qin Ma BMBL @QinMaBMBL]

Happy to announce our special issue on "Application of large language models in genome analysis", now live on @GenomeBiology. Honored to serve as a guest editor alongside @supercs08! We welcome your manuscript submissions to this groundbreaking collection.https://t.co/ICYZkYGhNb

[Jiangning (John) Song @supercs08]

@BiologyAIDaily Thanks for introducing our ImmuScope work. Very well summarized!

[Nature Methods @naturemethods]

Two absolutely fantastic bioimage analysis papers out today offering exceptional, generalizable tools for segmentation--Cellpose3 and Segment Anything for Microscopy. (1/3)

[tbepler @tbepler1]

Excited to share PoET-2, our next breakthrough in protein language modeling. It represents a fundamental shift in how AI learns from evolutionary sequences. 🧵 1/13

[Biology+AI Daily @BiologyAIDaily]

Self-iterative multiple instance learning enables the prediction of CD4+ T cell immunogenic epitopes

1. ImmuScope is an innovative deep-learning framework designed to predict CD4+ T cell immunogenic epitopes with high accuracy. It integrates multi-allelic and single-allelic data to enhance predictions in immune response initiation, making it a key tool for advancing vaccine and immunotherapy research.

2. The model leverages a self-iterative multiple-instance learning (MIL) approach combined with positive-anchor triplet loss, enabling it to handle weakly labeled multi-allelic data efficiently. This approach significantly increases the coverage of alleles while minimizing label ambiguity.

3. ImmuScope outperforms existing methods in multiple tasks, including antigen presentation prediction, MHC-II binding specificity discovery, CD4+ T cell epitope prediction, and immunogenicity assessment. Its robustness and generalizability were validated across diverse datasets, showing superior predictive power.

4. A key breakthrough of ImmuScope is its application in discovering melanoma neoantigens, where it revealed mutations in epitope sequences that affect pMHC-II binding and immunogenicity, providing deeper insights into cancer immunology.

5. The framework was also applied to SARS-CoV-2 epitope analysis, successfully predicting immune escape mechanisms in the Omicron variant and aligning with experimental data on immune response dynamics.

6. ImmuScope's ability to perform motif deconvolution on multi-allelic data provides critical insights into MHC-II binding specificities, aiding in the identification of new epitopes for immunotherapy and vaccine development.

7. This comprehensive solution not only enhances our understanding of T cell activation but also facilitates personalized medicine by identifying immunogenic neoepitopes and optimizing therapeutic interventions.

@supercs08 @RossjohnLab @Zachary_Zhikang @yumengzhang99
📜Paper: https://t.co/SY82I85HOr

#ImmuScope #TCellImmunity #Immunotherapy #VaccineDevelopment #MachineLearning #Immunopeptidomics #CancerImmunology #SARSCoV2 #Neoantigen

[Itai Yanai @ItaiYanai]

How to write a grant?
1. Write it for the reviewer, not you, the applicant.
2. Communicate in stories.
3. Make your story cohesive—leave no puzzling gaps.
4. Make your story resonate to keep the reviewer reading.
5. Accept chance and noise in peer-review.
https://t.co/wHZ065DNlm

[Jiangning (John) Song @supercs08]

- DrugDAGT accurately predicts synergistic drug combinations for SARS-CoV-2 treatment.

- We expect DrugDAGT to accelerate the discovery of safe and effective drug combinations for complex therapeutic applications.

@YaojiaChen0807

[Jiangning (John) Song @supercs08]

We introduce DrugDAGT, a new machine learning-based tool that leverages a dual-attention graph transformer framework with contrastive learning to predict multiple types of drug-drug interactions (DDIs). Work published in @BMC Biology

[Jiangning (John) Song @supercs08]

- We expect DrugDAGT to accelerate the discovery of safe and effective drug combinations for complex therapeutic applications.

Paper: https://t.co/VauyRsd55T

Code: https://t.co/2bHzHxVt5Z

[Jiangning (John) Song @supercs08]

- DrugDAGT effectively identifies key local structures critical to DDI outcomes, as evidenced in interactions involving ketoconazole and loxoprofen with various drugs.

- DrugDAGT accurately predicts synergistic drug combinations for SARS-CoV-2 treatment.

[Jiangning (John) Song @supercs08]

- DrugDAGT demonstrates significant improvements in DDI prediction for both warm-start and cold-start scenarios, leveraging the combined strengths of the graph transformer and contrastive learning frameworks.

[Jiangning (John) Song @supercs08]

- Fine-tuning EPACT on 3D data identifies binding hotspots within the TCR-pMHC complex and corroborates the presence of T cell cross-reactivity.

- EPACT is anticipated to accelerate the development of TCR-based immunotherapies and vaccines for infectious diseases and cancers.

[Jiangning (John) Song @supercs08]

Our latest work published @NatMachIntell introduces a bioinformatic tool to decipher the interaction between paired CD8+ T cell receptors and peptide-MHC complexes, and can predict for unseen epitopes.
@Jamie_Rossjohn @MarkGerstein

Paper: https://t.co/UIryLOxcFa

[Jiangning (John) Song @supercs08]

- It substantially improves the prediction of TCR binding specificity for unseen epitopes using the contrastive and transfer learning framework.
- The predictions for SARS-CoV-2 responsive T cells align well with the surge in spike-specific immune responses after vaccination.