Joshua Box (@DrJoshuaBox) profile picture

ML for bio and drug discovery. Recovering chemoinf. Occassionally FulhamFC and London. Views my own. Now @drjoshuabox.bsky.social

London
Joined April 2011
Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)Photo by Joshua Box (@DrJoshuaBox)
Joshua Box retweeted
Structure or sequence – how should we represent proteins? This question has long challenged SOTA protein property prediction. Read our new post to get our take and meet Boltz-2-PPI, a minimal adaptation of Boltz for predicting protein–protein affinity 🧬👇 open.substack.com/pub/synten…
1
2
1
159
Big congrats to @aaron_lou, @chenlin_meng & @StefanoErmon on winning an #ICML2024 Best Paper Award for: Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution Exciting work improving diffusion models for symbolic data like text. arxiv.org/abs/2310.16834
ICML 2024: Best Paper Awards: Florian Tramèr; Gautam Kamath; Nicholas Carlini. Position: Considerations for Differentially Private Learning with Large-Scale Public Pretraining Akbir Khan; John Hughes; Dan Valentine; Laura Ruis; Kshitij Sachan; Ansh Radhakrishnan; Edward Grefenstette; Samuel R. Bowman; Tim Rocktäschel; Ethan Perez. Debating with More Persuasive LLMs Leads to More Truthful Answers Jake Bruce; Michael D Dennis; Ashley Edwards; Jack Parker-Holder; Yuge Shi; Edward Hughes; Matthew Lai; Aditi Mavalankar; Richie Steigerwald; Chris Apps; Yusuf Aytar; Sarah Maria Elisabeth Bechtle; Feryal Behbahani; Stephanie C.Y. Chan; Nicolas Heess; Lucy Gonzalez; Simon Osindero; Sherjil Ozair; Scott Reed; Jingwei Zhang; Konrad Zolna; Jeff Clune; Nando de Freitas; Satinder Singh; Tim Rocktäschel. Genie: Generative Interactive Environments Dan Kondratyuk; Lijun Yu; Xiuye Gu; Jose Lezama; Jonathan Huang; Grant Schindler; Rachel Hornung; Vighnesh Birodkar; Jimmy Yan; Ming-Chang Chiu; Krishna Somandepalli; Hassan Akbari; Yair Alon; Yong Cheng; Joshua V. Dillon; Agrim Gupta; Meera Hahn; Anja Hauth; David Hendon; Alonso Martinez; David Minnen; Mikhail Sirotenko; Kihyuk Sohn; Xuan Yang; Hartwig Adam; Ming-Hsuan Yang; Irfan Essa; Huisheng Wang; David A Ross; Bryan Seybold; Lu Jiang. VideoPoet: A Large Language Model for Zero-Shot Video Generation Aaron Lou; Chenlin Meng; Stefano Ermon. Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution Patrick Esser; Sumith Kulal; Andreas Blattmann; Rahim Entezari; Jonas Müller; Harry Saini; Yam Levi; Dominik Lorenz; Axel Sauer; Frederic Boesel; Dustin Podell; Tim Dockhorn; Zion English; Robin Rombach. Scaling Rectified Flow Transformers for High-Resolution Image Synthesis Stephen Zhao; Rob Brekelmans; Alireza Makhzani; Roger Baker Grosse. Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo Dora Zhao; Jerone Andrews; Orestis Papakyriakopoulos; Alice Xiang. Measure Dataset Diversity, Don’t Just Claim It. Nicholas Carlini; Daniel Paleka; Krishnamurthy Dj Dvijotham; Thomas Steinke; Jonathan Hayase; A. Feder Cooper; Katherine Lee; Matthew Jagielski; Milad Nasr; Arthur Conmy; Eric Wallace; David Rolnick; Florian Tramèr. Stealing Part of a Production Language Model Idan Attias; Gintare Karolina Dziugaite; Mahdi Haghifam; Roi Livni; Daniel M. Roy. Information Complexity of Stochastic Convex Optimization: Applications to Generalization and Memorization
4
24
4
132
32,757
How do healthy tissues evolve to become cancerous? Delighted to share our latest research tracking the somatic evolution that occurs in the decades before AML. 👇Implications for early detection and early interception. biorxiv.org/cgi/content/shor…
3
67
9
240
36,030
Not an ML expert, but want to understand AlphaFold3 (AF3)? I wrote a simple-to-understand AF3 field guide and sprinkled my opinions in along the way. The full essay is linked at the END of the thread. Here's a TL;DR 🧵
1
88
8
337
65,601
New blog post alert! 🚨 I write about AlphaFold3, how it works, where it works well ✅ (and where we think it might not ❌) and what it means to the TechBio landscape more broadly (if you decide to believe me). 🧬 (1/7) 🧵 open.substack.com/pub/harris…
1
10
6
39
5,396
OpenFold paper is now out in @naturemethods. Timely. I’ll say more about AF3 at the end of this thread, but first I wanted to highlight what’s new since our initial release. (1/9)
9
257
10
1,059
184,870
We’re really happy to share our preprint demonstrating the atomically-accurate design of single-domain antibodies (VHHs) with RFdiffusion!
6
74
5
395
45,741
Joshua Box retweeted
here is sora, our video generation model: openai.com/sora today we are starting red-teaming and offering access to a limited number of creators. @_tim_brooks @billpeeb @model_mechanic are really incredible; amazing work by them and the team. remarkable moment.
1,470
3,901
2,393
24,064
6,250,114
The first Practical Cheminformatics post of the year is a retrospective of papers published in 2023, “AI in Drug Discovery 2023 - A Highly Opinionated Literature Review”. practicalcheminformatics.blo…
1
39
1
143
10,379
Use inverse folding likelihoods as an unsupervised protein sequence optimizer. @varunrshanker @theobruun @BrianHie Peter Kim
1
24
126
17,176
Joshua Box retweeted
some NeurIPS highlights: LDM tutorial was great. the 1st part is a must-watch, the 2nd part is full of references for recent advancements, and you need to check the 3rd part if you're planning to use LDMs for something other than images/videos. neurips2023-ldm-tutorial.git…
4
38
4
274
40,508
Ever since the invention of the car cities have been changing, and in the 1930s people like Norman Bel Geddes predicted exactly what would come to pass: cities designed for cars rather than people. A feeling captured rather well by this cartoon...
6
140
11
2,303
105,390
I got into ml because the other option was organic chemistry
you got into ML from a YouTube video, I got into ML to impress a girl, we are not the same
1
4
6
78
28,534
Joshua Box retweeted
New! We’ve just put up a note evaluating the latest, in-development version of AlphaFold (“AlphaFold-latest”). This is a preview - development is still in progress - but performance across a wide range of tasks is striking. deepmind.google/discover/blo… Highlights in the thread. 1/7
8
237
27
778
210,264
In 1938, Julian Layton was in Vienna helping to extract Jews from Austria – then the architect of the Final Solution rang. Here, our writer @JasperRees tells the story of his great uncle. Read the full article here: telegraph.co.uk/tv/2023/10/2…
7
5
2
24
12,669
Have you been seeking an alternative to LSTMs for de novo drug design? Have tried GPTs but couldn't find what you hoped for?😕 Look no more! We've just the thing for you: Structured State Space Sequence Models (S4) for de novo drug design🎉 A thread🧵(1/6)
2
8
2
54
14,044
One of the best ways to reduce hallucinations with LLMs is by retrieving useful, factual information and injecting it into the LLM’s prompt as added context. Although this might sound complicated, it’s actually quite easy to implement with standard vector search functionality… Why do we need this? All LLMs have a fixed context length. So, the amount of information we can include in a prompt is limited by nature! As such, we need to be selective about the context that we provide to our model. If we want to provide useful context that can reduce hallucinations and improve the model’s output, one of the best approaches is to retrieve relevant information from an external (vector) database. Retrieval framework for LLMs. Assuming that we have a lot of relevant textual data that can be used by an LLM, we can’t just inject all of this data into the model’s prompt every time that we perform inference. Rather, we need to do the following: 1. Break our data into textual chunks 2. Vectorize each of the textual chunks 3. Store these vectors (with their data) in a vector database 4. Find relevant data at inference time using vector search 5. Add relevant data to our prompt to provide more context to the LLM We will use the same embedding model to vectorize these chunks and to generate query vectors that can be used for search. Storing data in a vector db. The first step in the above framework is to chunk our data. Typically, we will use chunks of ~200 tokens. However, the optimal chunk size is a hyperparameter that can change depending on the application. Then, we use an embedding model to vectorize each of these chunks, and we can store them, along with their text data, in a vector database (e.g., Redis, Weaviate, Pinecone, Qdrant, etc.). Retrieving relevant context. When we want to retrieve relevant textual data from our vector db, we should just i) create a query embedding based on our prompt (possibly including the chat history) and ii) run a vector search for relevant documents. This way, we can use semantic search to identify portions of data that are relevant to include as context within the LLM’s prompt. Creating the query embedding. There are a ton of different ways we can create query embedding for searching our vector db. The simplest approach would be to truncate our chat history or prompt and pass this directly into the embedding model. But, if this is too long, we could ask the LLM to summarize our chat history or prompt before embedding it, or even to convert the chat history or prompt into a list of search keywords. Picking an embedding model. To make sure this works well, we need a good embedding model that captures the semantic similarities between our queries and textual chunks. There are a variety of good embedding models publicly available via SentenceTransformers and HuggingFace. To find one that works for you, I’d recommend taking a look at the Massive Text Embedding Benchmark (hosted on HuggingFace). The result. We can use the approach described above to power retrieval-augmented generation (RAG), which is one of the best ways to reduce hallucinations and improve the output of LLMs. Given that this approach can be implemented without significant effort via tools like Pinecone / Weaviate and HuggingFace / SentenceTransformers, it is undoubtedly one of the most useful practical tools for building with LLMs.
39
190
16
1,332
357,583
"From Proteins to Ligands: Decoding Deep Learning Methods for Binding Affinity Prediction" Found that ligand encodings, but not protein encodings, affect affinity prediction outcomes biorxiv.org/content/10.1101/… github.com/meyresearch/DL_pr…
1
52
2
265
23,965
The latest Practical Cheminformatics post, "We Need Better Benchmarks for Machine Learning in Drug Discovery", turned into a bit of a rant, but I had a lot I needed to say. #might_ruffle_some_feathers practicalcheminformatics.blo…
6
35
4
160
18,750
There's a new Practical Cheminformatics blog post, with lots of code, "Getting Real with Molecular Property Prediction". practicalcheminformatics.blo…
2
21
78
7,045