jinyuan sun

Interpretation of results

1

what does PROT_IDX_LIST indicate in the salt bridge? Also, the format is a little strange. For other records, the first 6 columns remain the same and in the salt bridge,...

File path in *.ds file

2

Hi, It seems the file name in `*.ds` file does not support an absolute path like `/home/user/dataset/*.pdb`. I have a `.ds` file: ``` /home/jsun/gpse_v2/test/structure_dataset/A0A1Q8LK65_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/W7ISS8_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/A5EDZ7_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/A0A3G7HEQ2_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/A0A6J4VWF3_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/A0A1R4K1N1_AF.pdb /home/jsun/gpse_v2/test/structure_dataset/A0A7W8E001_AF.pdb...

Finetune script

19

Could you provide a script/notebook demo to show how to finetune this model?

Add Fine-tuning Example

1

Addresses the discussion in issue #11 by introducing a new example script for fine-tuning the `evo-1-8k-base` model on the `human-genome-cds`.

The procedure of making the self-distillation data.

2

I'm curious about `.pdb` files your kindly provided for self-distillation. Were they predicted using official AF2 weights or the openfold initial training weights trained from scratch?

Hello, I am looking for guidance on how to design a pair of primers to amplify a specific plasmid sequence. I need to ensure that one primer extends downstream (forward...

HelixFold3 的测试数据集与AF3对齐的问题

1

速度很快的工作，很强。核酸的预测是很重要的问题，我也是关注者之一。HelixFold3提供了开源可用的工具，同时大幅度超越了RF2NA和RFAA，确实很棒。我有如下问题： 1. 想问一下CASP15的RNA测试的过程中，HelixFold3推理了几次？ 2. HelixFold3如何做的ranking？对RF2NA和RFAA，有多次采样重新排序么？ 3. AF3的指标是来源于AF3 server预测的结构自行计算的lddt么？如果是的话有更换seed重新ranking么？ 4. 另外RNA的recent pdb的测试数据集我观察到您们的效果也是很好的，但是我也有类似上面的疑问。

model not available on api.together.ai

3

Hi, I found the together API might be down again. The error I got is: Error code: 400 - {'error': {'message': 'Unable to access non-serverless model togethercomputer/evo-1-131k-base. Please visit https://api.together.ai/models/togethercomputer/evo-1-131k-base...

Can I run mmseqs easy-cluster with gapped sequences?

1

I would like to cluster sequences with gaps, can I use mmseqs linclust or easy-cluster? Will the gaps in input sequences affect the identity calculation?