Sabitlenmiş Tweet
I built a protein solubility prediction tool which runs on ESM-650M, fine tuned on UESolDS dataset.
Feel free to check it out.
github.com/61-Keys/oracle…
English
Asutosh Rath
1.5K posts
Asutosh Rath
@MusicalPlanet3
Independent ML researcher. I love tinkering around the Bio-AI domain.
Bengaluru, India Katılım Ağustos 2021
1.2K Takip Edilen108 Takipçiler
We work tirelessly to prevent a future that you don’t see
And if we’re successful
You won’t even know how close we got
To the blackpill
English
Someone keeps warning you
Of every development
Months before It occurs
There is only one way out;
Personal weights
Andy Fang@andyfang
Introducing Dasher Tasks Dashers can now get paid to do general tasks. We think this will be huge for building the frontier of physical intelligence. Look forward to seeing where this goes!
English
It’s a tragedy that most of you
are too retarded to accept it
simp 4 satoshi@iamgingertrash
When you find out About the non-human life, That walks amongst us; Will your mind accept it?
English
Asutosh Rath retweetledi
Rocket garden at Starbase.
You can see this from the public highway.
Dima Zeniuk@DimaZeniuk
SpaceX Starship rocket garden
English
@MusicalPlanet3 Who said nuking was the solution? High frequency weapons would be more than sufficient to incapacitate them.
English
But can it fight the Quiet place monster ?
K.L@kinglinzhuhui
China has developed an AI robot for this job, replacing humans in one of the most dangerous workplaces — grain storage facilities. These spaces are filled with dust, extreme heat, low oxygen levels, and unstable grain piles that can collapse without warning.
English
@shreyux but they would have to nuke the whole planet then. The monsters kept on coming.
English
@MusicalPlanet3 Realistically speaking the militaries of the world would destroy the aliens in a matter of weeks.
English
English
@Cherenedene @hobopony117 @elonmusk @inductionheads @beffjezos Try this;
Elon should think of a retarded idea, and pitch it to his team and present it convincingly
Tell them you really believe this is the path
Then fire everyone who agrees
Simple test
English
Fractional Cursor acquisition.
xAI taking the right steps
Jason Ginsberg@JasonBud
I’m proud to be joining SpaceX and xAI with @milichab It has become clear that software is changing fundamentally. More and more, people can shape the tools they use directly, and the ceiling of what can be built keeps rising. What makes xAI special is the scale of its ambition: to build from first principles all the way out to the stars. I’m especially grateful to work on products that expand human agency and freedom. That mission is deeply personal to me. My family came to the United States fleeing communism, and the belief that freedom should be part of the next generation of the internet has driven me every day since Andrew and I started Skiff. Now, we get to work on intelligence, understanding, and freedom on a universal scale.
English
@MusicalPlanet3 Oooo
Prime Intellect
Will
Karpathy
lots of names I like - thanks man
English
Asutosh Rath retweetledi
It would be perfect, if Anthropic collaborates with Apple and replaces Apple Intelligence with Claude.
English
@neural_avb github.com/61-Keys/AMP-MI…
I built a tool, which predicts the minimum inhibitory concentration on Antimicrobial peptides.
More info : x.com/MusicalPlanet3…
Asutosh Rath@MusicalPlanet3
I built AMP-MIC-Predictor : A model that predicts how potent antimicrobial peptides are against 16 different bacterial species, using ESM-2 embeddings. Pearson r = 0.828 | 43K data points | Open source github.com/61-Keys/AMP-MI…
English
Built with:
• ESM-2-650M (frozen embeddings)
• Learned pathogen embeddings (32-dim)
• MLP head (~1.1M params)
• Data from DBAASP via AMPGen
Previous work: I built ORACLE-Sol for protein solubility prediction. Now applying the same PLM-embedding approach to AMR.
English
I built AMP-MIC-Predictor : A model that predicts how potent antimicrobial peptides are against 16 different bacterial species, using ESM-2 embeddings.
Pearson r = 0.828 | 43K data points | Open source
github.com/61-Keys/AMP-MI…
English
The reason, we have less accuracy over different pathogens, is due to the availability of less data ; as compared to E coli and S. Aureus.
English
Per-pathogen results: E. coli → r = 0.87 S. aureus → r = 0.87 B. subtilis → r = 0.57 P. aeruginosa → r = 0.30
Performance tracks data availability. The unified architecture lets rare pathogens benefit from shared learning.
English
The key finding: pathogen conditioning matters.
ESM-2 + MLP (no pathogen info): Pearson 0.800 ESM-2 + Pathogen-Conditioned MLP: Pearson 0.828
Telling the model which bacterium you're targeting improves prediction. Different bacteria have different membrane compositions.
English
Most existing AMP predictors answer a simple yes/no: "is this an antimicrobial peptide?"
That's not useful in practice. A biologist needs to know: will THIS peptide kill THIS specific pathogen, and at what dose?
My model takes (peptide sequence + pathogen) → predicted MIC.
English