Biological databases in the age of generative artificial intelligence Abstract. Summary: Modern biological research critically depends on public databases. The introduction and propagation of errors within and across database

From #OUP @bioinfoadv.bsky.social journal, our paper is out, from discussions we had last summer at @iscb.bsky.social #ISMB2024 in #Montréal | Biological #databases in the age of generative artificial intelligence | #Bioinformatics #AI #LLM #OpenScience 🧬 🖥️ 🧪 🔓 
⬇️
academic.oup.com/bioinformati...

The publishing track talk I presented last year in Montreal #ISMB2024 on the challenges with peer-reviewers, framed using the medium of John Candy films, now seems to be online. Who wants an orange whip? www.youtube.com/watch?v=meck...

24. Yennamalli lab winning Wiki competition
25. Completed a 4 month long science writing workshop
26. Writing reports on recent publications for Current Science and sharpening my sci-comm skills
27. Conducting a special session on publishing research at #ISMB2024, Montreal

Conference poster image

My conference poster from ISMB 2024 is now publicly available on ResearchGate:
dx.doi.org/10.13140/RG....
It has a DOI and clickable links — feel free to explore and even cite it🧪
#ISMB2024

Conference presentation of "Disparate radiomic imaging features are predictive of recurrence events and molecular subtype in Black and White breast cancer patients"

Thank you #ISMB2024 for letting us share our collaboration with Dr. Acquaah-Mensah of MCPHS Boston where we used AI/ML and ISB-CGC to analyze radiomic data in Black and White breast cancer patients. We were excited by the enthusiasm for using Generative AI in #CancerResearch!

Our write-up of last week's birthday celebrations and conference goings in Montreal.

#ISMB2024 and #BOSC2024 – Demystifying the World of AI (and Publishing)
gigasciencejournal.com/blog/ismb202...

Finally en route home ✈️🌏 - Merci pour tout Montréal 🇨🇦 & #ISMB2024

2024 ISCB Innovator Award recipient, Dr. Sui-In Lee, concludes the #ISMB2024 congress with a powerful message: the researchers' goal is not to publish in prestigious journals, but to push the boundaries of human knowledge.

SL: Concept-based explanations; explainable AI for foundation models, generative AI; AI auditing for biomedicine #ISMB2024

SL: Future directions: there are so many things to explain! In biology and clinical medicine, same underlying AI principle is applied. Future of medicine requires the convergence of all modalities. Would like to see more researchers working on both. #ISMB2024

SL: Explainable AI involves complex computer science challenges and have demonstrated significant impact in medicine, healthcare, cancer biology and precision medicine. #ISMB2024

SL: Advancing SHAP and explainable AI. Unifying principle: explaining by removing features #ISMB2024

SL: Importance of explainable AI #ISMB2024 www.nature.com/articles/s42...

Now: Su-In Lee: Explainable AI for health: where we are and how to move forward #ISMB2024 💻🧬🧪

JD starts talk with a slide listing a couple of dozen author-journal name citations, and then another listing presentations given by lab members earlier in the conference. #ISMB2024

Now: Jishnu Das: SLIDE: Significant Latent Factor Interaction Discovery and Exploration across biological domains #ISMB2024 💻🧬🧪

AI: PathFA provides interpretable representations on pathway level on multiple modalities, balances multimodal information with different noise levels, efficient and works with few samples. #ISMB2024

AI: Paper #ISMB2024 academic.oup.com/bioinformatics/article/4...

AI: Experiments from simulated data using MSigDB HallMark gene sets and associations and sample B from mixture. Unimodal setting: more data efficient than PLIER. Standard factor analysis does not recover ground truth. Multimodal: Proteomic view profits from RNA data but not vice versa. #ISMB2024

AI: Empirical Bayes (ML-II) on hyperparameters: max_{Λ,δ,_rna,σ_prot} log p(Y_rna,YΛ,…) #ISMB2024

AI: Inference with alternating least-squares with amortized empirical Bayes. maximum a posteriori: max_{U,B} log p(Y_rna, Y_prot, U, B|Λ,δ,σ_rna,σ_prot) #ISMB2024

AI: Probabilistic model of data matrix Y_{genes × samples}, shared latent space U_{gene sets × latents}, B_{latents × samples} per marker m and sample s: p(Y|U,B,σ) = Π_{m,s} N(Y_{m,s}|cₘ^TUbₛ,σₘ² #ISMB2024

AI: Pathway-based multimodal factor analysis. Need to choose gene sets depending on the problem such as MSigDB cell type or MSigDB Hallmark annotations. → binary matrix C of gene set marker annotations #ISMB2024

AI: Integrating multiple modalities into sample representations. Examples of bulk RNA and proteotyping, validated with CyTOF. #ISMB2024

Now: Alexander Immer: PathFA: probabilistic pathway-based multimodal factor analysis #ISMB2024 💻🧬🧪

Now: Mor Nitzan: Towards spatiotemporal design principles in multicellular systems #ISMB2024 💻🧬🧪

Thomas Lengauer explains the behind the doors of publishing in Bioinformatics Advances @BioinfoAdv, the open access journal of the International Society for Computational Biology. #ISMB2024

Alex Bateman discusses fair and unfair uses of generative AI and LLMs in scientific publishing at #ISMB2024. Unmissable if one want to use GTP for assistance.

PTM-Mamba: A PTM-Aware Protein Language Model with Bidirectional Gated Mamba Blocks Proteins serve as the workhorses of living organisms, orchestrating a wide array of vital functions. Post-translational modifications (PTMs) of their amino acids greatly influence the structural and functional diversity of different protein types and uphold proteostasis, allowing cells to swiftly respond to environmental changes and intricately regulate complex biological processes. To this point, efforts to model the complex features of proteins have involved the training of large and expressive protein language models (pLMs) such as ESM-2 and ProtT5, which accurately encode structural, functional, and physicochemical properties of input protein sequences. However, the over 200 million sequences that these pLMs were trained on merely scratch the surface of proteomic diversity, as they neither input nor account for the effects of PTMs. In this work, we fill this major gap in protein sequence modeling by introducing PTM tokens into the pLM training regime. We then leverage recent advancements in structured state space models (SSMs), specifically Mamba, which utilizes efficient hardware-aware primitives to overcome the quadratic time complexities of Transformers. After adding a comprehensive set of PTM tokens to the model vocabulary, we train bidirectional Mamba blocks whose outputs are fused with state-of-the-art ESM-2 embeddings via a novel gating mechanism. We demonstrate that our resultant PTM-aware pLM, PTM-Mamba , improves upon ESM-2’s performance on various PTM-specific tasks. PTM-Mamba is the first and only pLM that can uniquely input and represent both wild-type and PTM sequences, motivating downstream modeling and design applications specific to post-translationally modified proteins. To facilitate PTM-aware protein language modeling applications, we have made our model available at: <https://huggingface.co/ChatterjeeLab/PTM-Mamba>. ### Competing Interest Statement The authors have declared no competing interest.

Now: Zhangzhi Peng: PTM-Mamba: a PTM-aware protein language model with bidirectional gated Mamba blocks #ISMB2024 💻🧬🧪

JR: Editing genome of HAP1 cells with sgRNAs that target genes. #ISMB2024