feed.xml

<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.3.4">Jekyll</generator><link href="https://zitniklab.hms.harvard.edu/feed.xml" rel="self" type="application/atom+xml" /><link href="https://zitniklab.hms.harvard.edu/" rel="alternate" type="text/html" /><updated>2025-02-10T22:44:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/feed.xml</id><title type="html">Zitnik Lab</title><subtitle>Harvard Machine Learning for Medicine and Science</subtitle><author><name>Marinka Zitnik</name></author><entry><title type="html">MedTok: Unlocking Medical Codes for GenAI</title><link href="https://zitniklab.hms.harvard.edu/2025/02/10/MedTok/" rel="alternate" type="text/html" title="MedTok: Unlocking Medical Codes for GenAI" /><published>2025-02-10T00:00:00-05:00</published><updated>2025-02-10T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/02/10/MedTok</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/02/10/MedTok/"><![CDATA[<p>Meet <a href="https://arxiv.org/pdf/2502.04397">MedTok, a multimodal medical code tokenizer that transforms how AI understands structured medical data.</a> By integrating textual descriptions and relational contexts, MedTok enhances tokenization for transformer-based models—powering everything from EHR foundation models to medical QA.</p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[Meet MedTok, a multimodal medical code tokenizer that transforms how AI understands structured medical data. By integrating textual descriptions and relational contexts, MedTok enhances tokenization for transformer-based models—powering everything from EHR foundation models to medical QA.]]></summary></entry><entry><title type="html">What If You Could Rewrite Biology? Meet CLEF</title><link href="https://zitniklab.hms.harvard.edu/2025/02/10/CLEF/" rel="alternate" type="text/html" title="What If You Could Rewrite Biology? Meet CLEF" /><published>2025-02-10T00:00:00-05:00</published><updated>2025-02-10T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/02/10/CLEF</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/02/10/CLEF/"><![CDATA[<p>What if we could anticipate molecular and medical changes before they happen? Introducing <a href="https://arxiv.org/pdf/2502.03569">CLEF, an approach for counterfactual generation in biological and medical sequence models.</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[What if we could anticipate molecular and medical changes before they happen? Introducing CLEF, an approach for counterfactual generation in biological and medical sequence models.]]></summary></entry><entry><title type="html">Digital Twins as Global Health and Disease Models</title><link href="https://zitniklab.hms.harvard.edu/2025/02/05/DigitalTwins/" rel="alternate" type="text/html" title="Digital Twins as Global Health and Disease Models" /><published>2025-02-05T00:00:00-05:00</published><updated>2025-02-05T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/02/05/DigitalTwins</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/02/05/DigitalTwins/"><![CDATA[<p>New paper on the role of digital twins as <a href="https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-025-01435-7">global health and disease learning models for preventive and personalized medicine.</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[New paper on the role of digital twins as global health and disease learning models for preventive and personalized medicine.]]></summary></entry><entry><title type="html">LLM and KG+LLM agent papers at ICLR</title><link href="https://zitniklab.hms.harvard.edu/2025/01/26/ICLRpapers/" rel="alternate" type="text/html" title="LLM and KG+LLM agent papers at ICLR" /><published>2025-01-26T00:00:00-05:00</published><updated>2025-01-26T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/01/26/ICLRpapers</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/01/26/ICLRpapers/"><![CDATA[<p>New papers on <a href="https://arxiv.org/abs/2407.06483">test-time interventions in language models</a> and <a href="https://arxiv.org/abs/2410.04660">knowledge graph based LLM agents</a> accepted to ICLR. <a href="https://zitniklab.hms.harvard.edu/projects/KGARevion/">[KGARevion]</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[New papers on test-time interventions in language models and knowledge graph based LLM agents accepted to ICLR. [KGARevion]]]></summary></entry><entry><title type="html">Artificial Intelligence in Medicine 2</title><link href="https://zitniklab.hms.harvard.edu/2025/01/25/AIM2/" rel="alternate" type="text/html" title="Artificial Intelligence in Medicine 2" /><published>2025-01-25T00:00:00-05:00</published><updated>2025-01-25T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/01/25/AIM2</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/01/25/AIM2/"><![CDATA[<p>Excited to share our new graduate course on <a href="https://zitniklab.hms.harvard.edu/AIM2/">Artificial Intelligence in Medicine 2</a>.</p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[Excited to share our new graduate course on Artificial Intelligence in Medicine 2.]]></summary></entry><entry><title type="html">AI Design of Proteins for Therapeutics</title><link href="https://zitniklab.hms.harvard.edu/2025/01/05/CellSystemsVoices/" rel="alternate" type="text/html" title="AI Design of Proteins for Therapeutics" /><published>2025-01-05T00:00:00-05:00</published><updated>2025-01-05T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/01/05/CellSystemsVoices</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/01/05/CellSystemsVoices/"><![CDATA[<p>New Voices piece in Cell Systems: <a href="https://www.cell.com/cell-systems/abstract/S2405-4712(24)00309-0">How will computational protein design change biotechnology and therapeutic development?</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[New Voices piece in Cell Systems: How will computational protein design change biotechnology and therapeutic development?]]></summary></entry><entry><title type="html">ProCyon AI Highlighted by Kempner</title><link href="https://zitniklab.hms.harvard.edu/2025/01/05/ProCyonKempnerBlog/" rel="alternate" type="text/html" title="ProCyon AI Highlighted by Kempner" /><published>2025-01-05T00:00:00-05:00</published><updated>2025-01-05T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2025/01/05/ProCyonKempnerBlog</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2025/01/05/ProCyonKempnerBlog/"><![CDATA[<p>Thanks to Kempner Institute for highlighting our latest research, <a href="https://kempnerinstitute.harvard.edu/research/deeper-learning/procyon-a-multimodal-foundation-model-for-protein-phenotypes/">ProCyon, our protein-text foundation model for modeling protein functions.</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[Thanks to Kempner Institute for highlighting our latest research, ProCyon, our protein-text foundation model for modeling protein functions.]]></summary></entry><entry><title type="html">Foundation Model for Protein Phenotypes</title><link href="https://zitniklab.hms.harvard.edu/2024/12/16/ProCyon/" rel="alternate" type="text/html" title="Foundation Model for Protein Phenotypes" /><published>2024-12-16T00:00:00-05:00</published><updated>2024-12-16T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2024/12/16/ProCyon</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2024/12/16/ProCyon/"><![CDATA[<p>New paper: <a href="https://www.biorxiv.org/content/10.1101/2024.12.10.627665v1">ProCyon is a groundbreaking foundation model for modeling, generating, and predicting protein phenotypes</a>. <a href="https://zitniklab.hms.harvard.edu/ProCyon/">[Project website]</a> <a href="https://github.com/mims-harvard/ProCyon">[Code]</a></p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[New paper: ProCyon is a groundbreaking foundation model for modeling, generating, and predicting protein phenotypes. [Project website] [Code]]]></summary></entry><entry><title type="html">SPECTRA in Nature Machine Intelligence</title><link href="https://zitniklab.hms.harvard.edu/2024/12/07/SPECTRA/" rel="alternate" type="text/html" title="SPECTRA in Nature Machine Intelligence" /><published>2024-12-07T00:00:00-05:00</published><updated>2024-12-07T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2024/12/07/SPECTRA</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2024/12/07/SPECTRA/"><![CDATA[<p>Are biomedical AI models truly as smart as they seem? <a href="https://www.nature.com/articles/s42256-024-00931-6">SPECTRA is a framework that evaluates models by considering the full spectrum of cross-split overlap: train-test similarity.</a> SPECTRA reveals gaps in benchmarks for molecular sequence data across 19 models, including LLMs, GNNs, diffusion models, and conv nets.</p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[Are biomedical AI models truly as smart as they seem? SPECTRA is a framework that evaluates models by considering the full spectrum of cross-split overlap: train-test similarity. SPECTRA reveals gaps in benchmarks for molecular sequence data across 19 models, including LLMs, GNNs, diffusion models, and conv nets.]]></summary></entry><entry><title type="html">Unified Clinical Vocabulary Embeddings</title><link href="https://zitniklab.hms.harvard.edu/2024/12/07/UnifiedClinicalVocabularyEmbeddings/" rel="alternate" type="text/html" title="Unified Clinical Vocabulary Embeddings" /><published>2024-12-07T00:00:00-05:00</published><updated>2024-12-07T00:00:00-05:00</updated><id>https://zitniklab.hms.harvard.edu/2024/12/07/UnifiedClinicalVocabularyEmbeddings</id><content type="html" xml:base="https://zitniklab.hms.harvard.edu/2024/12/07/UnifiedClinicalVocabularyEmbeddings/"><![CDATA[<p>New paper: <a href="https://www.medrxiv.org/content/10.1101/2024.12.03.24318322">A unified resource provides a new representation of clinical knowledge by unifying medical vocabularies.</a> (1) Phenotype risk score analysis across 4.57 million patients, (2) Inter-institutional clinician panels evaluate alignment with clinical knowledge across 90 diseases and 3,000 clinical codes.</p>]]></content><author><name>Marinka Zitnik</name></author><summary type="html"><![CDATA[New paper: A unified resource provides a new representation of clinical knowledge by unifying medical vocabularies. (1) Phenotype risk score analysis across 4.57 million patients, (2) Inter-institutional clinician panels evaluate alignment with clinical knowledge across 90 diseases and 3,000 clinical codes.]]></summary></entry></feed>