Foldseek

https://github.com/steineggerlab/foldseek
Admin
Foldseek enables fast and sensitive comparisons of large protein structure sets, supporting monomer...
Tokens:773,715
Snippets:4,683
Trust Score:7.4
License:GPL-3.0
Update:2 weeks ago
Show doc for...
Tokens:
Raw
### Install Corrosion from source

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/setup_corrosion.md

Use these commands to download and install Corrosion from source. Optionally specify -DCMAKE_INSTALL_PREFIX to set a custom installation directory. The install step may require sudo or admin privileges when installing to a system location.

```bash
git clone https://github.com/corrosion-rs/corrosion.git
# Optionally, specify -DCMAKE_INSTALL_PREFIX=<target-install-path> to specify a 
# custom installation directory
cmake -Scorrosion -Bbuild -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release
# This next step may require sudo or admin privileges if you're installing to a system location,
# which is the default.
cmake --install build --config Release
```

--------------------------------

### Install cargo-instruments and profile with MacOS Instruments

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/README.md

Installs cargo-instruments via Homebrew and runs the profile example with release optimizations and AVX2 SIMD features, opening the Instruments GUI.

```bash
brew install cargo-instruments
RUSTFLAGS="-g" cargo instruments --example profile --release --features simd_avx2 --open
```

--------------------------------

### Build and install zstd-adaptive

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/contrib/adaptive-compression/README.md

Build the adaptive compression tool in the adaptive-compression directory under contrib, or install it as the command zstd-adaptive.

```bash
make adapt
```

```bash
make install
```

--------------------------------

### Install google-cloud-cpp via vcpkg

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Installs the google-cloud-cpp package from vcpkg, required for Google Cloud Storage support in createdb.

```bash
git clone https://github.com/microsoft/vcpkg.git
./vcpkg/bootstrap-vcpkg.sh
./vcpkg/vcpkg install google-cloud-cpp
```

--------------------------------

### Build and install zstandard with Meson

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/build/meson/README.md

Navigate to the meson directory and run these commands to configure, build, and install zstandard. The build type is release and contrib and tests are enabled.

```sh
meson --buildtype=release -D with-contrib=true -D with-tests=true -D with-contrib=true builddir
cd builddir
ninja             # to build
ninja install     # to install
```

--------------------------------

### Install Foldseek on Linux with SSE4.1

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Download and install the statically compiled SSE4.1 version for Linux.

```bash
wget https://mmseqs.com/foldseek/foldseek-linux-sse41.tar.gz
tar xvzf foldseek-linux-sse41.tar.gz
export PATH=$(pwd)/foldseek/bin/:$PATH
```

--------------------------------

### Run pssm_accuracy example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the pssm_accuracy example with AVX2 SIMD features in release mode and captures the output.

```python
output = !cd .. && cargo run --example pssm_accuracy --release --features simd_avx2 --quiet
output
```

--------------------------------

### Install to staging directory with DESTDIR

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/build/meson/README.md

Use DESTDIR to install zstandard into a staging directory, useful for packaging or testing without affecting the system.

```sh
DESTDIR=./staging ninja install
```

--------------------------------

### Install Foldseek on Linux with AVX2

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Download and install the statically compiled AVX2 version for Linux.

```bash
wget https://mmseqs.com/foldseek/foldseek-linux-avx2.tar.gz
tar xvzf foldseek-linux-avx2.tar.gz
export PATH=$(pwd)/foldseek/bin/:$PATH
```

--------------------------------

### Run uc_bench benchmark with simd_avx2

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_bench_vis.ipynb

Runs the uc_bench example from the block-aligner crate with the simd_avx2 feature in release mode, capturing the command output into a variable. The output is a list of strings starting with '# time (s)' and containing benchmark results for algorithms 'ours (no trace)', 'ours (trace)', and 'parasail' on the uc30 dataset.

```python
output = !cd .. && cargo run --example uc_bench --release --features simd_avx2 --quiet
output
```

--------------------------------

### Install Rust target standard library

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/usage.md

Use rustup to install the standard library for the target Rust triple before cross-compiling.

```bash
rustup target add <target-rust-triple>
```

--------------------------------

### Run block_img example to generate images

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the block_img example from the block-aligner crate with AVX2 SIMD features in release mode. Generates two PNG images (block_img1.png and block_img2.png) in the vis directory. The command is executed from the notebook's parent directory.

```bash
!cd .. && cargo run --example block_img --release --features simd_avx2 --quiet -- vis/block_img1.png vis/block_img2.png
```

--------------------------------

### Run block-aligner compare example on adaptivebandbench scores

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the compare example from the block-aligner crate for each adaptivebandbench score file, using the simd_avx2 feature and a band width of 100000. The output is a list of comparison results.

```python
path_prefix = "../adaptivebandbench/"
outputs = []
for f in output:
    o = !cd .. && cargo run --example compare --release --features simd_avx2 --quiet -- {path_prefix + f} 100000
    outputs.append(o)
outputs
```

--------------------------------

### Install Foldseek on macOS

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Download and install the universal macOS build with SSE4.1/AVX2/M1 NEON support.

```bash
wget https://mmseqs.com/foldseek/foldseek-osx-universal.tar.gz
tar xvzf foldseek-osx-universal.tar.gz
export PATH=$(pwd)/foldseek/bin/:$PATH
```

--------------------------------

### Install cross-compiler for Linux PowerPC64LE

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/usage.md

On Ubuntu, install g++-powerpc64le-linux-gnu to cross-compile for 64-bit Little-Endian PowerPC.

```bash
sudo apt install g++-powerpc64le-linux-gnu
```

--------------------------------

### printf formatting example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/fmt/README.md

Shows the concise printf equivalent for the same formatting task, highlighting the verbosity difference between iostreams and printf.

```c
printf("%.2f\n", 1.23456);
```

--------------------------------

### Run block-aligner compare example on diff-bench-paper scores

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the compare example from the block-aligner crate for each diff-bench-paper score file, using the simd_avx2 feature and a band width of 50. The output shows comparison statistics for each file.

```python
path_prefix = "../diff-bench-paper/"
outputs = []
for f in output:
    o = !cd .. && cargo run --example compare --release --features simd_avx2 --quiet -- {path_prefix + f} 50
    outputs.append(o)
outputs
```

```text
Result:
[['max size, total, other better, other % better, us better, us % better',
  '',
  '32, 1734, 21, 0.08564959576734898, 1652, 0.014478023753787836',
  '',
  '64, 1734, 6, 0.014165306396410624, 1669, 0.017846773911076242',
  '# Done!'],
 ['max size, total, other better, other % better, us better, us % better',
  '',
  '32, 1734, 123, 0.11277759543387586, 1546, 0.0018081439985594904',
  '',
  '64, 1734, 56, 0.01591056402384434, 1633, 0.01032615586212946',
  '# Done!'],
 ['max size, total, other better, other % better, us better, us % better',
  '',
  '32, 1734, 203, 0.1292233687300956, 892, 0.0016197787311747142',
  '',
  '64, 1734, 75, 0.013415647980847437, 1051, 0.02590382915635615',
  '# Done!'],
 ['max size, total, other better, other % better, us better, us % better',
  '',
  '32, 1734, 229, 0.1193901258355045, 37, 0.024771049421762822',
  '',
  '64, 1734, 86, 0.013091483866668505, 176, 0.1557904770879331',
  '# Done!']]
```

--------------------------------

### Install Foldseek on Linux with GPU support

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Download and install the GPU-enabled version for Linux, which supports ProstT5.

```bash
wget https://mmseqs.com/foldseek/foldseek-linux-gpu.tar.gz
tar xvzf foldseek-linux-gpu.tar.gz
export PATH=$(pwd)/foldseek/bin/:$PATH
```

--------------------------------

### Example output with zlib

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/zlibWrapper/README.md

Output of the example.c program compiled with zlib only, showing successful decompression and gz functions.

```text
zlib version 1.2.8 = 0x1280, compile flags = 0x65
uncompress(): hello, hello!
gzread(): hello, hello!
gzgets() after gzseek:  hello!
inflate(): hello, hello!
large_inflate(): OK
after inflateSync(): hello, hello!
inflate with dictionary: hello, hello!
```

--------------------------------

### Install Foldcomp and download binaries

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/foldcomp/README.md

Install the Foldcomp Python package with pip and download static binaries for Linux, Linux ARM64, macOS, and Windows.

```bash
# Install Foldcomp Python package
pip install foldcomp

# Download static binaries for Linux
wget https://mmseqs.com/foldcomp/foldcomp-linux-x86_64.tar.gz

# Download static binaries for Linux (ARM64)
wget https://mmseqs.com/foldcomp/foldcomp-linux-arm64.tar.gz

# Download binary for macOS
wget https://mmseqs.com/foldcomp/foldcomp-macos-universal.tar.gz

# Download binary for Windows (x64)
wget https://mmseqs.com/foldcomp/foldcomp-windows-x64.zip
```

--------------------------------

### Benchmark with custom input arguments

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/contrib/experimental_dict_builders/benchmarkDictBuilder/README.md

Run the benchmark with custom input arguments using make ARG= followed by the permitted arguments. The example uses two input directories.

```makefile
make ARG="in=../../../lib/dictBuilder in=../../../lib/compress"
```

--------------------------------

### Download AlphaFoldDB Swiss-Prot with foldcomp.setup

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/foldcomp/README.md

Downloads the AlphaFoldDB Swiss-Prot database using the foldcomp Python helper. The command is run from the command line and requires the foldcomp Python package to be installed.

```bash
python -c "import foldcomp; foldcomp.setup('afdb_swissprot_v4');"
```

--------------------------------

### Run accuracy example and capture output

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the accuracy example from the block-aligner crate with AVX2 SIMD features in release mode and captures its output into the variable 'output'. The output is a list of strings containing CSV-like accuracy results. This is the first step in the accuracy analysis workflow.

```python
output = !cd .. && cargo run --example accuracy --release --features simd_avx2 --quiet
output
```

--------------------------------

### Install MMseqs2 via brew, conda, docker, or precompiled binaries

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/README.md

Install MMseqs2 via Homebrew, conda, Docker, or by downloading a precompiled binary. The binary variants include GPU (requires glibc >= 2.29 and nvidia driver >=525.60.13), AVX2 (fastest), AVX2-64bit (supports >2^32 sequences but needs 2x more memory), SSE4.1, SSE2 (slowest), and macOS universal.

```bash
brew install mmseqs2
```

```bash
conda install -c conda-forge -c bioconda mmseqs2
```

```bash
docker pull ghcr.io/soedinglab/mmseqs2
```

```bash
wget https://mmseqs.com/latest/mmseqs-linux-gpu.tar.gz; tar xvfz mmseqs-linux-gpu.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

```bash
wget https://mmseqs.com/latest/mmseqs-linux-avx2.tar.gz; tar xvfz mmseqs-linux-avx2.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

```bash
wget https://mmseqs.com/latest/mmseqs-linux-avx2-64bit.tar.gz; tar xvfz mmseqs-linux-avx2-64bit.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

```bash
wget https://mmseqs.com/latest/mmseqs-linux-sse41.tar.gz; tar xvfz mmseqs-linux-sse41.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

```bash
wget https://mmseqs.com/latest/mmseqs-linux-sse2.tar.gz; tar xvfz mmseqs-linux-sse2.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

```bash
wget https://mmseqs.com/latest/mmseqs-osx-universal.tar.gz; tar xvfz mmseqs-osx-universal.tar.gz; export PATH=$(pwd)/mmseqs/bin/:$PATH
```

--------------------------------

### Example output with zstd wrapper

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/zlibWrapper/README.md

Output after switching to zstd_zlibwrapper.h with -DZWRAP_USE_ZSTD=1, with test_flush and test_sync disabled.

```text
zlib version 1.2.8 = 0x1280, compile flags = 0x65
uncompress(): hello, hello!
gzread(): hello, hello!
gzgets() after gzseek:  hello!
inflate(): hello, hello!
large_inflate(): OK
inflate with dictionary: hello, hello!
```

--------------------------------

### Install Foldseek via wget, tar, and PATH export

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Install Foldseek using precompiled binaries for Linux (AVX2, ARM64, GPU) and macOS, or via Conda. Each command downloads the archive, extracts it, and exports the binary path. The GPU build requires glibc >=2.17 and nvidia driver >=525.60.13.

```bash
# Linux AVX2 build (check using: cat /proc/cpuinfo | grep avx2)
wget https://mmseqs.com/foldseek/foldseek-linux-avx2.tar.gz; tar xvzf foldseek-linux-avx2.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

```

```bash
# Linux ARM64 build
wget https://mmseqs.com/foldseek/foldseek-linux-arm64.tar.gz; tar xvzf foldseek-linux-arm64.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

```

```bash
# Linux AVX2 & GPU build (req. glibc >= 2.17 and nvidia driver >=525.60.13)
wget https://mmseqs.com/foldseek/foldseek-linux-gpu.tar.gz; tar xvfz foldseek-linux-gpu.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

```

```bash
# MacOS
wget https://mmseqs.com/foldseek/foldseek-osx-universal.tar.gz; tar xvzf foldseek-osx-universal.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

```

```bash
# Conda installer (Linux and macOS)
conda install -c conda-forge -c bioconda foldseek

```

--------------------------------

### Run nanopore_accuracy example with cargo run

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the nanopore_accuracy example with AVX2 SIMD features and captures its output. The output is a list of strings containing CSV-like summary lines and sequence identity averages.

```python
output = !cd .. && cargo run --example nanopore_accuracy --release --features simd_avx2 --quiet
output
```

```text
Result:
['',
 'dataset, total, wrong, wrong % error, min size wrong, wfa wrong',
 '',
 'illumina, 100000, 0, NaN, 0, 0',
 '# illumina seq id avg: 0.997052517030022',
 '',
 'nanopore 1kbp, 12477, 21, 0.060149406709727474, 1810, 21',
 '# nanopore 1kbp seq id avg: 0.8926079817605514',
 '',
 'nanopore <10kbp, 5000, 109, 0.035888133253518265, 1595, 645',
 '# nanopore <10kbp seq id avg: 0.8752513435635519',
 '',
 'nanopore <50kbp, 10000, 278, 0.020799640807527112, 2599, 2545',
 '# nanopore <50kbp seq id avg: 0.8795798677370729',
 '# Done!']
```

--------------------------------

### Define and install structty_render static library

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/structty/src/render/CMakeLists.txt

Defines the static library structty_render from Renderer.cpp and AnsiOutput.cpp, sets public include directories using generator expressions for both build-tree and install-tree contexts, and installs the library, headers, and CMake package files. The comment notes that no ncurses linkage is intentional and suggests verifying with `nm -u libstructty_render.a | grep -i ncurses`.

```cmake
add_library(structty_render STATIC
    Renderer.cpp
    AnsiOutput.cpp
)

# Use generator expressions so include paths are correct in both build-tree
# (add_subdirectory) and install-tree (find_package) contexts.
target_include_directories(structty_render
    PUBLIC
        $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include>
        $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
        $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/src/visualization>
        $<INSTALL_INTERFACE:include>
)

# No ncurses linkage — intentional; consumers verify with:
#   nm -u libstructty_render.a | grep -i ncurses
# ── Install rules ──────────────────────────────────────────────────────────────
include(GNUInstallDirs)

install(TARGETS structty_render
    EXPORT  structty_renderTargets
    ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
)

# Public headers: structty_render.h + render headers + RenderPoint (used by both)
install(FILES
    ${PROJECT_SOURCE_DIR}/include/structty_render.h
    ${CMAKE_CURRENT_SOURCE_DIR}/Renderer.hpp
    ${CMAKE_CURRENT_SOURCE_DIR}/AnsiOutput.hpp
    ${PROJECT_SOURCE_DIR}/src/visualization/RenderPoint.hpp
    DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
)

# CMake package files for find_package(structty_render) support
install(EXPORT structty_renderTargets
    FILE      structty_renderTargets.cmake
    NAMESPACE structty_render::
    DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/structty_render
)

install(FILES
    ${PROJECT_SOURCE_DIR}/cmake/structty_renderConfig.cmake
    DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/structty_render
)
```

--------------------------------

### Representative FASTA output (_repseq.fasta)

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Example of the _repseq.fasta output: FASTA sequences of cluster representative proteins.

```fasta
>Q0KJ32
MAGA....R
>E3HQM9
MCAT...Q
```

--------------------------------

### Build with simd_wasm for wasm32-wasi

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/README.md

Build the repository with WASM SIMD support. To run the resulting WASM programs, you need wasmtime installed and on your $PATH.

```bash
cargo build --target=wasm32-wasi --features simd_wasm --release
```

--------------------------------

### ZSTD_compressBegin_usingDict

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/doc/zstd_manual.html

Start a compression operation using a dictionary.

```APIDOC
## ZSTD_compressBegin_usingDict

### Description
Start a compression operation using a dictionary.

### Method
Function

### Signature
```c
size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
```

### Parameters
- **cctx** (ZSTD_CCtx*) - Required - The compression context.
- **dict** (const void*) - Required - The dictionary to use.
- **dictSize** (size_t) - Required - The size of the dictionary.

### Return Value
- Size_t indicating success or error.
```

--------------------------------

### Align sequences with ksw_extz and print CIGAR

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/ksw2/README.md

Complete example of using KSW2 to align two sequences. It encodes the sequences to 0/1/2/3, calls ksw_extz with a 5x5 scoring matrix, and prints the resulting CIGAR. The example uses match score 1, mismatch -2, gap open 2, and gap extension 1.

```c
#include <string.h>
#include <stdio.h>
#include "ksw2.h"

void align(const char *tseq, const char *qseq, int sc_mch, int sc_mis, int gapo, int gape)
{
	int i, a = sc_mch, b = sc_mis < 0? sc_mis : -sc_mis; // a>0 and b<0
	int8_t mat[25] = { a,b,b,b,0, b,a,b,b,0, b,b,a,b,0, b,b,b,a,0, 0,0,0,0,0 };
	int tl = strlen(tseq), ql = strlen(qseq);
	uint8_t *ts, *qs, c[256];
	ksw_extz_t ez;

	memset(&ez, 0, sizeof(ksw_extz_t));
	memset(c, 4, 256);
	c['A'] = c['a'] = 0; c['C'] = c['c'] = 1;
	c['G'] = c['g'] = 2; c['T'] = c['t'] = 3; // build the encoding table
	ts = (uint8_t*)malloc(tl);
	qs = (uint8_t*)malloc(ql);
	for (i = 0; i < tl; ++i) ts[i] = c[(uint8_t)tseq[i]]; // encode to 0/1/2/3
	for (i = 0; i < ql; ++i) qs[i] = c[(uint8_t)qseq[i]];
	ksw_extz(0, ql, qs, tl, ts, 5, mat, gapo, gape, -1, -1, 0, &ez);
	for (i = 0; i < ez.n_cigar; ++i) // print CIGAR
		printf("%d%c", ez.cigar[i]>>4, "MID"[ez.cigar[i]&0xf]);
	putchar('\n');
	free(ez.cigar); free(ts); free(qs);
}

int main(int argc, char *argv[])
{
	align("ATAGCTAGCTAGCAT", "AGCTAcCGCAT", 1, -2, 2, 1);
	return 0;
}
```

--------------------------------

### Run uc_accuracy example and capture output

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/block-aligner/vis/block_aligner_accuracy_vis.ipynb

Runs the uc_accuracy example from the block-aligner crate with AVX2 SIMD features and captures its output. The output is a list of strings containing accuracy statistics for the Uniclust30 dataset at various sequence identity thresholds and block sizes, including wrong counts and error percentages.

```python
output = !cd .. && cargo run --example uc_accuracy --release --features simd_avx2 --quiet
output
```

```text
Result:
['# seq identity is lower bound (inclusive)',
 'dataset, size, seq identity, count, wrong, wrong % error',
 'uc30_0.95, 32-32, 0, 0, 0, NaN',
 'uc30_0.95, 32-32, 0.1, 0, 0, NaN',
 'uc30_0.95, 32-32, 0.2, 14, 0, NaN',
 'uc30_0.95, 32-32, 0.3, 873, 45, 0.24234488792701261',
 'uc30_0.95, 32-32, 0.4, 1166, 70, 0.23740147684159452',
 'uc30_0.95, 32-32, 0.5, 951, 39, 0.22458752668094342',
 'uc30_0.95, 32-32, 0.6, 923, 30, 0.2423639248577639',
 'uc30_0.95, 32-32, 0.7, 789, 19, 0.23070399690337745',
 'uc30_0.95, 32-32, 0.8, 747, 9, 0.2314009291547539',
 'uc30_0.95, 32-32, 0.9, 1537, 18, 0.14197227628986994',
 '',
 '# total: 7000, wrong: 230, wrong % error: 0.22858669521170225, length avg: 329.554, length min: 22, length max: 8881, dp fraction: 0.3389349961353417',
 '',
 'uc30_0.95, 32-256, 0, 0, 0, NaN',
 'uc30_0.95, 32-256, 0.1, 0, 0, NaN',
 'uc30_0.95, 32-256, 0.2, 14, 0, NaN',
 'uc30_0.95, 32-256, 0.3, 873, 10, 0.022419704677810615',
 'uc30_0.95, 32-256, 0.4, 1166, 11, 0.0640908027337777',
 'uc30_0.95, 32-256, 0.5, 951, 5, 0.06678752297229766',
 'uc30_0.95, 32-256, 0.6, 923, 9, 0.04391515678016153',
 'uc30_0.95, 32-256, 0.7, 789, 4, 0.023700475250152283',
 'uc30_0.95, 32-256, 0.8, 747, 1, 0.0006338028169014084',
 'uc30_0.95, 32-256, 0.9, 1537, 4, 0.05817779039335834',
 '',
 '# total: 7000, wrong: 44, wrong % error: 0.04514810836644425, length avg: 329.554, length min: 22, length max: 8881, dp fraction: 0.3654793322322297',
 '',
 'uc30_0.95, 256-256, 0, 0, 0, NaN',
 'uc30_0.95, 256-256, 0.1, 0, 0, NaN',
 'uc30_0.95, 256-256, 0.2, 14, 0, NaN',
 'uc30_0.95, 256-256, 0.3, 873, 1, 0.029182542431589884',
 'uc30_0.95, 256-256, 0.4, 1166, 2, 0.015912013783823824',
 'uc30_0.95, 256-256, 0.5, 951, 0, NaN',
 'uc30_0.95, 256-256, 0.6, 923, 1, 0.14847403884264765',
 'uc30_0.95, 256-256, 0.7, 789, 0, NaN',
 'uc30_0.95, 256-256, 0.8, 747, 0, NaN',
 'uc30_0.95, 256-256, 0.9, 1537, 0, NaN',
 '',
 '# total: 7000, wrong: 4, wrong % error: 0.052370152210471296, length avg: 329.554, length min: 22, length max: 8881, dp fraction: 4.221035489298381',
 '',
 'uc30, 32-32, 0, 0, 0, NaN',
 'uc30, 32-32, 0.1, 0, 0, NaN',
 'uc30, 32-32, 0.2, 192, 82, 0.9275327360430471',
 'uc30, 32-32, 0.3, 1170, 288, 0.5938895288958876',
 'uc30, 32-32, 0.4, 1182, 272, 0.5563610451866828',
 'uc30, 32-32, 0.5, 1160, 229, 0.5338615587300842',
 'uc30, 32-32, 0.6, 1044, 192, 0.5196576365562074',
 'uc30, 32-32, 0.7, 847, 103, 0.4959494251346888',
 'uc30, 32-32, 0.8, 694, 86, 0.47644760063092073',
 'uc30, 32-32, 0.9, 711, 19, 0.25554500500876537',
 '',
 '# total: 7000, wrong: 1271, wrong % error: 0.564413277478051, length avg: 302.51342857142856, length min: 25, length max: 8293, dp fraction: 0.35536869224678297',
 '',
 'uc30, 32-256, 0, 0, 0, NaN',
 'uc30, 32-256, 0.1, 0, 0, NaN',
 'uc30, 32-256, 0.2, 192, 12, 0.14227202149727955',
 'uc30, 32-256, 0.3, 1170, 49, 0.12495803143160221',
 'uc30, 32-256, 0.4, 1182, 49, 0.12113074192604727',
 'uc30, 32-256, 0.5, 1160, 35, 0.15550386498366273',
 'uc30, 32-256, 0.6, 1044, 32, 0.11973947097066927',
 'uc30, 32-256, 0.7, 847, 18, 0.06996993682039082',
 'uc30, 32-256, 0.8, 694, 21, 0.08632626264571577',
 'uc30, 32-256, 0.9, 711, 8, 0.045885200156420294',
 '',
 '# total: 7000, wrong: 224, wrong % error: 0.11821118070339384, length avg: 302.51342857142856, length min: 25, length max: 8293, dp fraction: 0.44600432314797367',
 '',
 'uc30, 256-256, 0, 0, 0, NaN',
 'uc30, 256-256, 0.1, 0, 0, NaN',
 'uc30, 256-256, 0.2, 192, 0, NaN',
 'uc30, 256-256, 0.3, 1170, 7, 0.3835458905002951',
 'uc30, 256-256, 0.4, 1182, 4, 0.7080088654647194',
 'uc30, 256-256, 0.5, 1160, 7, 0.4602842002549759',
 'uc30, 256-256, 0.6, 1044, 5, 0.3023075188513019',
 'uc30, 256-256, 0.7, 847, 2, 0.23043440955768113',
 'uc30, 256-256, 0.8, 694, 1, 0.0544986745672852',
 'uc30, 256-256, 0.9, 711, 1, 0.0690854119425548',
 '',
 '# total: 7000, wrong: 27, wrong % error: 0.4012902443343513, length avg: 302.51342857142856, length min: 25, length max: 8293, dp fraction: 4.391380248306236',
 '',
 '# Done!']
```

--------------------------------

### Advanced compression options example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/programs/zstd.1.md

Sets advanced compression options similar to predefined level 19 for files larger than 256 KB. The parameters include window log, chain log, hash log, search log, search length, target length, and strategy.

```text
--zstd=wlog=23,clog=23,hlog=22,slog=6,slen=3,tlen=48,strat=6
```

--------------------------------

### ZSTD_compressBegin_advanced

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/doc/zstd_manual.html

Start a compression operation with finer parameter control.

```APIDOC
## ZSTD_compressBegin_advanced

### Description
Start a compression operation with finer parameter control.

### Method
Function

### Signature
```c
size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_parameters params, unsigned long long pledgedSrcSize);
```

### Parameters
- **cctx** (ZSTD_CCtx*) - Required - The compression context.
- **dict** (const void*) - Required - The dictionary to use.
- **dictSize** (size_t) - Required - The size of the dictionary.
- **params** (ZSTD_parameters) - Required - Compression parameters.
- **pledgedSrcSize** (unsigned long long) - Required - Pledged source size.

### Return Value
- Size_t indicating success or error.
```

--------------------------------

### Compile Foldseek from source under Linux

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Clone, build, and install Foldseek from source. Requires git, g++ 4.9+, and cmake 2.8.12+. The binary will be in build/bin/.

```bash
git clone https://github.com/steineggerlab/foldseek.git
cd foldseek
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=RELEASE -DCMAKE_INSTALL_PREFIX=. ..
make
make install 
export PATH=$(pwd)/bin/:$PATH
```

--------------------------------

### IPS⁴o usage example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/ips4o/README.md

Shows how to include the IPS⁴o header and use the sequential and parallel sort functions. The parallel version uses OpenMP if available, otherwise std::thread. Requires C++14 and a CPU supporting 16-byte compare-and-exchange.

```C++
#include "ips4o.hpp"

// sort sequentially
ips4o::sort(begin, end[, comparator])

// sort in parallel (uses OpenMP if available, std::thread otherwise)
ips4o::parallel::sort(begin, end[, comparator])
```

--------------------------------

### All member FASTA output (_allseq.fasta)

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Example of the _allseq.fasta output. A new cluster is marked by two identical name lines of the representative sequence, followed by the FASTA-formatted sequences of all its members.

```fasta
>Q0KJ32	
>Q0KJ32
MAGA....R
>C0W539
MVGA....R
>D6KVP9
MVGA....R
>E3HQM9	
>E3HQM9
MCAT...Q
>Q223C0
MCAR...Q
```

--------------------------------

### ZSTD_compressBegin

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/doc/zstd_manual.html

Start a compression operation. This is part of the buffer-less streaming compression API.

```APIDOC
## ZSTD_compressBegin

### Description
Start a compression operation. This is part of the buffer-less streaming compression API.

### Method
Function

### Signature
```c
size_t ZSTD_compressBegin(ZSTD_CCtx* cctx);
```

### Parameters
- **cctx** (ZSTD_CCtx*) - Required - The compression context.

### Return Value
- Size_t indicating success or error.
```

--------------------------------

### iostreams formatting example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/fmt/README.md

Shows the verbose iostreams approach to formatting a floating-point number with fixed precision, which the author contrasts with the simpler printf syntax.

```c++
std::cout << std::setprecision(2) << std::fixed << 1.23456 << "\n";
```

--------------------------------

### Compile and generate zstd manual

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/contrib/gen_html/README.md

Commands to compile the gen_html program and generate the zstd manual. The example uses version 1.1.1 and the header file ../../lib/zstd.h to produce zstd_manual.html.

```bash
make
./gen_html.exe 1.1.1 ../../lib/zstd.h zstd_manual.html
```

--------------------------------

### rust2cpp project file structure

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/usage.md

File structure for the rust2cpp example project, showing the layout of the Rust crate and the CMake project.

```text
rust2cpp/
    rust/
        src/
            lib.rs
        Cargo.lock
        Cargo.toml
    CMakeLists.txt
    main.cpp
```

--------------------------------

### ZSTD_resetCStream

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/doc/zstd_manual.html

Starts a new compression job using the same parameters from the previous job.

```APIDOC
## ZSTD_resetCStream

### Description
Starts a new compression job using the same parameters from the previous job. This is useful to skip dictionary loading since the dictionary is re-used in-place. The stream must be initialized at least once before calling this function.

### Signature
```c
size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize);
```

### Parameters
- **zcs** (ZSTD_CStream*) - Required - Pointer to the compression stream to reset.
- **pledgedSrcSize** (unsigned long long) - Required - The expected source size. If unknown, use `ZSTD_CONTENTSIZE_UNKNOWN`. If > 0, must be correct as it will be written in the header and controlled at the end. For compatibility, 0 is interpreted as "srcSize unknown" but will change to mean "empty" in future versions.

### Returns
- **size_t** - 0 on success, or an error code (testable with `ZSTD_isError()`).
```

--------------------------------

### Example representative multimer FASTA output

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Example of the representative multimer FASTA output file `_rep_seq.fasta` from easy-multimercluster. Each representative multimer is prefixed with a '#' comment line, followed by FASTA entries for each chain (header starting with '>' and sequence line). The content is truncated with '...' as shown in the source.

```plaintext
#5o002
>5o002_A
SHGK...R
>5o002_B
SHGK...R
#194l2
>194l2_A0
KVFG...L
>194l2_A6
KVFG...L
#10mh121
...
```

--------------------------------

### Help command

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/contrib/pzstd/README.md

Shows how to display the help information for PZstandard, which includes the smart default number of threads.

```bash
pzstd --help
```

--------------------------------

### Install macOS Clang dependencies with Homebrew

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Installs required dependencies for compiling Foldseek with Clang on macOS. Requires Xcode or Command Line Tools and libomp, recommended to install via Homebrew.

```bash
brew install cmake libomp zlib bzip2
```

--------------------------------

### Install Foldseek via bioconda

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Install Foldseek using conda from conda-forge and bioconda channels.

```bash
conda install -c conda-forge -c bioconda foldseek
```

--------------------------------

### Memory requirement example for AFDB50

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Example calculation for AFDB50 dataset with full memory usage.

```text
8 byte * 54*10^6 (Seqs) * 350 (avg. protein length) = 151G
```

--------------------------------

### Create custom database and index, then search with foldseek

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Pre-processes a target database with 'createdb' for repeated searches against the same structures. Optionally creates and stores the index with 'createindex'. Then runs a search with 'easy-search'.

```bash
foldseek createdb example/ targetDB
foldseek createindex targetDB tmp  #OPTIONAL generates and stores the index on disk
foldseek easy-search example/d1asha_ targetDB aln.m8 tmpFolder
```

--------------------------------

### Clone and configure format-benchmark repository

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/fmt/README.md

Commands to clone the format-benchmark repository and generate Makefiles with CMake, as a prerequisite for running the benchmarks.

```bash
$ git clone --recursive https://github.com/fmtlib/format-benchmark.git
$ cd format-benchmark
$ cmake .
```

--------------------------------

### Install macOS GCC dependencies with Homebrew

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Installs required packages for compiling Foldseek with GCC on macOS using Homebrew.

```bash
brew install cmake gcc@11 zlib bzip2
```

--------------------------------

### Memory requirement example without Cα

Source: https://github.com/steineggerlab/foldseek/wiki/Home

Example calculation for AFDB50 when using --sort-by-structure-bits 0, reducing memory to 35 GB.

```text
2 bytes * 54 x 10^6 (sequences) * 350 (average protein length) = 35 GB
```

--------------------------------

### Find Corrosion package in CMakeLists.txt

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/setup_corrosion.md

After installing Corrosion and ensuring the install directory is in PATH or CMAKE_PREFIX_PATH, use this command in your CMakeLists.txt to use Corrosion like any other package.

```cmake
find_package(Corrosion REQUIRED)
```

--------------------------------

### Create database with makedb

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/libmarv/Readme.md

Creates a database from a FASTA file (optionally gzip'ed). Supports up to 2 billion sequences. The output database is stored in dbfolder/dbname.

```bash
mkdir -p dbfolder
./makedb input.fa(.gz) dbfolder/dbname [options]
```

--------------------------------

### te_interp usage example

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/tinyexpr/README.md

Example usage of te_interp. Passing 0 as the error pointer ignores errors; passing a pointer sets it to 0 on success or the parse error position on failure.

```C
    int error;

    double a = te_interp("(5+5)", 0); /* Returns 10. */
    double b = te_interp("(5+5)", &error); /* Returns 10, error is set to 0. */
    double c = te_interp("(5+5", &error); /* Returns NaN, error is set to 4. */
```

--------------------------------

### Configure Meson build options

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/build/meson/README.md

Run meson configure to view or modify build options for the zstandard build.

```sh
meson configure
```

--------------------------------

### Equivalent commands for easy-interfacesearch

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Equivalent commands for interface-level search: the first runs `easy-interfacesearch` directly, while the second shows the step-by-step process it performs internally (createdb, createdimerdb, createinterfacedb, easy-multimersearch). Both are provided as equivalent ways to achieve the same result. The source states the commands below are equivalent.

```plaintext
foldseek easy-interfacesearch example example result tmpFolder
```

```plaintext
foldseek createdb example/ db
foldseek createdimerdb db dimerdb dimertmp
foldseek createinterfacedb dimerdb interfacedb
foldseek easy-multimersearch interfacedb interfacedb result tmpFolder
```

--------------------------------

### Example tab-separated multimercluster output

Source: https://github.com/steineggerlab/foldseek/blob/master/README.md

Example of the tab-separated output file `_cluster.tsv` from easy-multimercluster. The file maps each representative multimer (first column) to its member multimers (second column). The exact whitespace and formatting are as shown in the output.

```plaintext
5o002	   5o002
194l2	   194l2
194l2	   193l2
10mh121	 10mh121
10mh121	 10mh114
10mh121	 10mh119
```

--------------------------------

### Build both architectures with VS2015

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/build/VS_scripts/README.md

Runs the build script to compile both Release Win32 and Release x64 versions. The output is placed in the corresponding bin\Release\{ARCH}\ folder.

```batch
build.VS2015.cmd
```

--------------------------------

### Build both architectures with VS2013

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/zstd/build/VS_scripts/README.md

Runs the build script to compile both Release Win32 and Release x64 versions. The output is placed in the corresponding bin\Release\{ARCH}\ folder.

```batch
build.VS2013.cmd
```

--------------------------------

### Install Corrosion via Homebrew

Source: https://github.com/steineggerlab/foldseek/blob/master/lib/mmseqs/lib/corrosion/doc/src/setup_corrosion.md

Use this command to install Corrosion via Homebrew. Note that this package is community maintained and Corrosion follows semantic versioning; minor version bumps may contain breaking changes while pre-1.0.

```bash
brew install corrosion
```
Tokens:
Raw