HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
facebookresearch
esm
Python2023–2023
facebookresearch / esm · Archived

esm

Evolutionary Scale Modeling (esm): Pretrained language models for proteins

2Distinct people
3Mentions
0Explicit recommenders
4.2KGitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

Evolutionary Scale Modeling (esm): Pretrained language models for proteins

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 9–Nov 16 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/16/2026, 7:27:27 AM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by week3 dated mentions

Observed mentions: 2023-03-31 06:56:42 UTC to 2023-08-07 22:15:10 UTC. Weeks start Monday, UTC.

012023-03-27 00:00:00 UTC to 2023-04-03 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsMar 272023-04-03 00:00:00 UTC to 2023-04-10 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-04-10 00:00:00 UTC to 2023-04-17 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-04-17 00:00:00 UTC to 2023-04-24 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-04-24 00:00:00 UTC to 2023-05-01 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-05-01 00:00:00 UTC to 2023-05-08 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-05-08 00:00:00 UTC to 2023-05-15 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 comments2023-05-15 00:00:00 UTC to 2023-05-22 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-05-22 00:00:00 UTC to 2023-05-29 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-05-29 00:00:00 UTC to 2023-06-05 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 commentsMay 292023-06-05 00:00:00 UTC to 2023-06-12 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-06-12 00:00:00 UTC to 2023-06-19 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-06-19 00:00:00 UTC to 2023-06-26 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-06-26 00:00:00 UTC to 2023-07-03 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-07-03 00:00:00 UTC to 2023-07-10 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-07-10 00:00:00 UTC to 2023-07-17 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-07-17 00:00:00 UTC to 2023-07-24 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-07-24 00:00:00 UTC to 2023-07-31 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-07-31 00:00:00 UTC to 2023-08-07 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2023-08-07 00:00:00 UTC to 2023-08-14 00:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsAug 7

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Mar 272023-03-27 00:00:00 UTC2023-04-03 00:00:00 UTC101
Apr 32023-04-03 00:00:00 UTC2023-04-10 00:00:00 UTC000
Apr 102023-04-10 00:00:00 UTC2023-04-17 00:00:00 UTC000
Apr 172023-04-17 00:00:00 UTC2023-04-24 00:00:00 UTC000
Apr 242023-04-24 00:00:00 UTC2023-05-01 00:00:00 UTC000
May 12023-05-01 00:00:00 UTC2023-05-08 00:00:00 UTC000
May 82023-05-08 00:00:00 UTC2023-05-15 00:00:00 UTC101
May 152023-05-15 00:00:00 UTC2023-05-22 00:00:00 UTC000
May 222023-05-22 00:00:00 UTC2023-05-29 00:00:00 UTC000
May 292023-05-29 00:00:00 UTC2023-06-05 00:00:00 UTC000
Jun 52023-06-05 00:00:00 UTC2023-06-12 00:00:00 UTC000
Jun 122023-06-12 00:00:00 UTC2023-06-19 00:00:00 UTC000
Jun 192023-06-19 00:00:00 UTC2023-06-26 00:00:00 UTC000
Jun 262023-06-26 00:00:00 UTC2023-07-03 00:00:00 UTC000
Jul 32023-07-03 00:00:00 UTC2023-07-10 00:00:00 UTC000
Jul 102023-07-10 00:00:00 UTC2023-07-17 00:00:00 UTC000
Jul 172023-07-17 00:00:00 UTC2023-07-24 00:00:00 UTC000
Jul 242023-07-24 00:00:00 UTC2023-07-31 00:00:00 UTC000
Jul 312023-07-31 00:00:00 UTC2023-08-07 00:00:00 UTC000
Aug 72023-08-07 00:00:00 UTC2023-08-14 00:00:00 UTC101

What Hacker News says

1 sources
When evaluating this work, it’s important to remember that the functional labels and protein family assignments on each of the 280 million input sequences were originally assigned by an HMM model using human curated sequence groups as part of the pfam project, so the model is predicting a prediction (or perhaps conditioned on a prediction would be more accurate). Furthermore, the authors must engage a lot of human curation to ensure the sequences they generate are active. First, they pick an easy target. Second, they employ by-hand classical bioinformatics techniques on their predicted sequences after they are generated. For example, they manually align them and select those which contain specific important amino acids at specific positions which are present in 100% of functional proteins of that class, and are required for function. This
Read the rest
is all done by a human bioinformatics expert (or automated) before they test the generated sequences. This is the protein equivalent of cherry-picking great examples of, for example, ChatGPT responses and presenting them as if the model only made predictions like that. One other comment, in protein science, a sequence with 40% identity to another sequence is not “very different” if it is homologous. Since this model is essentially generating homologs from a particular class, it’s no surprise at a pairwise amino acid level, the generated sequences have this degree of similarity. Take proteins in any functional family and compare them. They will have the same overall 3-D structure—called their “fold”—yet have pairwise sequence identities much lower than 30–40%. This “degeneracy”, the notion that there are many diverse sequences that all fold into the same shape, is both a fundamental empirical observation in protein science as well as a grounded physical theory. Not to be negative. I really enjoyed reading this paper and I think the work is important. Some related work by Meta AI is the ESM series of models [1] trained on the same data (the UniProt dataset [2]). One thing I wonder is about the vocabulary size of this model. The number of tokens is 26 for the 20 amino acids and some extras, whereas for a LLM like Meta’s LLaMa the vocab size is 32,000. I wonder how that changes training and inference, and how we can adopt the transformer architecture for this scenario. 1. https://github.com/facebookresearch/esm 2. https://www.uniprot.org/help/downloads
In Large language models generate functional protein sequences across families

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.