Protocol to predict amino acid probabilities using the ESM language model family. More...

#include <EsmPerplexityTensorflowProtocol.hh>

Inheritance diagram for protocols::esm_perplexity::EsmPerplexityTensorflowProtocol:

Public Member Functions
	EsmPerplexityTensorflowProtocol ()=default
	Default constructor. More...

	EsmPerplexityTensorflowProtocol (EsmPerplexityTensorflowProtocol const &)=default
	Copy constructor. More...

	~EsmPerplexityTensorflowProtocol () override=default
	Destructor. More...

std::string	name () const override
	Get the name of this protocol. More...

	EsmPerplexityTensorflowProtocol (std::string const &model)
	constructor with name of the model More...

basic::tensorflow_manager::RosettaTensorflowProtocolBaseOP	clone () const override
	Clone operation: make a copy of this object, and return an owning pointer to the copy. More...

void	set_defaults ()
	set default options More...

bool	auto_download () const
	get auto_download value More...

void	auto_download (bool setting)
	set the auto_download value More...

Static Public Member Functions
static basic::citation_manager::CitationCollectionBaseCOP	get_ESM_neural_net_citation ()
	Get the citation for ESM. More...

static void	softmax (std::map< core::Size, utility::vector1< core::Real >> const &logit_map, std::map< core::Size, utility::vector1< core::Real >> &softmax_map)
	The tensorflow session. More...

static void	download_model_if_not_existing (std::string const &path_to_model, std::string const &base_url, bool const &auto_download)
	Downloads model from GitLab if the specified path does not exist or is empty. More...

Public Attributes
std::string	model_name_
	path to the model More...

Static Public Attributes
static std::string const	alphabet_ = "XXXXLAGVSERTIDPKQNFYMHWC"
	The ESM alphabet used for tokenization. More...

Static Protected Member Functions
static utility::vector1< core::Size >	get_selected_residues (core::select::residue_selector::ResidueSubset const &selected)
	Given a selection, return a vector if selected indices. More...

Static Private Member Functions
static bool	is_model_valid (std::string const &model_name)
	Checks if the given model name is one of the available models. More...

static std::string	concatenate_model_names ()
	Returns a string containing the names of all available models, separated by commas. More...

Private Attributes
bool	auto_download_ = false

Static Private Attributes
static std::string const	base_url_ = "https://git.iwe-lab.de/moritzertelt/ML_graphs/-/archive/main/ML_graphs-main.tar.gz?path=tensorflow_graphs/ESM/"

static std::vector< std::string > const	available_models_ = {"esm2_t6_8M_UR50D", "esm2_t12_35M_UR50D", "esm2_t30_150M_UR50D", "esm2_t33_650M_UR50D"}

Detailed Description

Protocol to predict amino acid probabilities using the ESM language model family.

Author: Moritz Ertelt (morit.nosp@m.z.er.nosp@m.telt@.nosp@m.goog.nosp@m.lemai.nosp@m.l.co.nosp@m.m)

Constructor & Destructor Documentation

◆ EsmPerplexityTensorflowProtocol() [1/3]

protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::EsmPerplexityTensorflowProtocol ( )

default

Default constructor.

◆ EsmPerplexityTensorflowProtocol() [2/3]

protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::EsmPerplexityTensorflowProtocol ( EsmPerplexityTensorflowProtocol const & )

default

Copy constructor.

◆ ~EsmPerplexityTensorflowProtocol()

protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::~EsmPerplexityTensorflowProtocol ( )

overridedefault

Destructor.

◆ EsmPerplexityTensorflowProtocol() [3/3]

protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::EsmPerplexityTensorflowProtocol ( std::string const & model )

explicit

constructor with name of the model

Constructor using the model name to initialize the particular ESM model.

References set_defaults().

Member Function Documentation

◆ auto_download() [1/2]

bool protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::auto_download ( ) const

inline

get auto_download value

References auto_download_.

Referenced by download_model_if_not_existing(), and set_defaults().

◆ auto_download() [2/2]

void protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::auto_download ( bool setting )

set the auto_download value

Set the auto_download bool.

References auto_download_.

◆ clone()

basic::tensorflow_manager::RosettaTensorflowProtocolBaseOP protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::clone ( ) const

override

Clone operation: make a copy of this object, and return an owning pointer to the copy.

◆ concatenate_model_names()

std::string protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::concatenate_model_names ( )

staticprivate

Returns a string containing the names of all available models, separated by commas.

References available_models_.

◆ download_model_if_not_existing()

void protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::download_model_if_not_existing	(	std::string const &	path_to_model,
		std::string const &	base_url,
		bool const &	auto_download
	)

static

Downloads model from GitLab if the specified path does not exist or is empty.

Downloads model from GitLab if the specified path does not exist or is missing crucial files.

Parameters

[in]	path_to_model	Directory path where the model should be located
[in]	base_url	Url of the GitLab where models are stored
[in]	auto_download	Whether to automatically download missing models

References auto_download(), and protocols::TR().

◆ get_ESM_neural_net_citation()

basic::citation_manager::CitationCollectionBaseCOP protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::get_ESM_neural_net_citation ( )

static

Get the citation for ESM.

Referenced by protocols::esm_perplexity::PerResidueEsmProbabilitiesMetric::provide_citation_info().

◆ get_selected_residues()

utility::vector1< core::Size > protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::get_selected_residues ( core::select::residue_selector::ResidueSubset const & selected )

staticprotected

Given a selection, return a vector if selected indices.

Given a selection, return the number of selected residues.

◆ is_model_valid()

bool protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::is_model_valid ( std::string const & model_name )

staticprivate

Checks if the given model name is one of the available models.

References available_models_.

◆ name()

std::string protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::name ( ) const

override

Get the name of this protocol.

◆ set_defaults()

void protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::set_defaults ( )

set default options

References auto_download().

Referenced by EsmPerplexityTensorflowProtocol().

◆ softmax()

void protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::softmax	(	std::map< core::Size, utility::vector1< core::Real >> const &	logit_map,
		std::map< core::Size, utility::vector1< core::Real >> &	softmax_map
	)

static

The tensorflow session.

Function to return the softmax of logits resulting in probabilities.

Apply softmax function to logits

Parameters

[in]	logit_map	A map of logits, where each key represents a residue and its predicted logits
[in,out]	softmax_map	A map of softmax, where each key represents a residue and its predicted probabilities.

References core::scoring::epr_deer::logit(), and core::scoring::pair.

Referenced by protocols::esm_perplexity::PerResidueEsmProbabilitiesMetric::calculate().

Member Data Documentation

◆ alphabet_

std::string const protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::alphabet_ = "XXXXLAGVSERTIDPKQNFYMHWC"

static

The ESM alphabet used for tokenization.

Referenced by protocols::esm_perplexity::PerResidueEsmProbabilitiesMetric::fill_return_map().

◆ auto_download_

bool protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::auto_download_ = false

private

Referenced by auto_download().

◆ available_models_

std::vector< std::string > const protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::available_models_ = {"esm2_t6_8M_UR50D", "esm2_t12_35M_UR50D", "esm2_t30_150M_UR50D", "esm2_t33_650M_UR50D"}

staticprivate

Referenced by concatenate_model_names(), and is_model_valid().

◆ base_url_

std::string const protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::base_url_ = "https://git.iwe-lab.de/moritzertelt/ML_graphs/-/archive/main/ML_graphs-main.tar.gz?path=tensorflow_graphs/ESM/"

staticprivate

◆ model_name_

std::string protocols::esm_perplexity::EsmPerplexityTensorflowProtocol::model_name_

path to the model

The documentation for this class was generated from the following files:

src/protocols/esm_perplexity/EsmPerplexityTensorflowProtocol.hh
src/protocols/esm_perplexity/EsmPerplexityTensorflowProtocol.cc

Public Member Functions

Static Public Member Functions

Public Attributes

Static Public Attributes

Static Protected Member Functions

Static Private Member Functions

Private Attributes

Static Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ EsmPerplexityTensorflowProtocol() [1/3]

◆ EsmPerplexityTensorflowProtocol() [2/3]

◆ ~EsmPerplexityTensorflowProtocol()

◆ EsmPerplexityTensorflowProtocol() [3/3]

Member Function Documentation

◆ auto_download() [1/2]

◆ auto_download() [2/2]

◆ clone()

◆ concatenate_model_names()

◆ download_model_if_not_existing()

◆ get_ESM_neural_net_citation()

◆ get_selected_residues()

◆ is_model_valid()

◆ name()

◆ set_defaults()

◆ softmax()

Member Data Documentation

◆ alphabet_

◆ auto_download_

◆ available_models_

◆ base_url_

◆ model_name_