Python Module Documentation¶

Recommended usage¶

In general, we recommend using metapredict in Python by first importing metapredict as meta:

import metapredict as meta

The meta module can the be used to call all the user-facing functions. Documentation for these functions is included below.

metapredict functions¶

metapredict.print_metapredict_legacy_network_version()[source]¶

Function that returns a string with the current trained network version used in disorder prediction. This is useful to know if updated versions of the network are provided, which will always accompany a version bump so prior versions of the code will always be available.

Returns:	Returns a string in the format v<version information>
Return type:	str

metapredict.print_metapredict_network_version()[source]¶

Function that returns a string with the current trained network version used in disorder prediction. This is useful to know if updated versions of the network are provided, which will always accompany a version bump so prior versions of the code will always be available.

Returns:	Returns a string in the format v<version information>
Return type:	str

metapredict.print_performance(seq_len=500, num_seqs=100, verbose=True, legacy=False)[source]¶

Function that lets you test metapredicts performance on your local hardware.

Parameters:	seqlen (int) – Length of each random sequence to be tested. Default = 500. num_seqs (int) – Number of sequences to compute over. Default = 100. verbose (bool) – Flag which, if true, means the function prints a summary when finished. If false simply returns an integer legacy (bool) – Flag which determines if legacy (v1) or updated (v2) metapredict networks are used.
Returns:	Returns the nearest number of sequences-per-second metapredict is currently predicting. For ref, on a spring 2020 MBP this value was ~10,000 sequences per second.
Return type:	int

metapredict.meta.predict_disorder_domains(sequence, disorder_threshold=None, minimum_IDR_size=12, minimum_folded_domain=50, gap_closure=10, normalized=True, return_numpy=True, legacy=False, return_list=False)[source]¶

This function takes an amino acid sequence and one or more variable options and returns a data structure called a DisorderObject. The object parameters associated with this object are defined below.

The previous version of metapredict returned a list of values, which can be obtained instead of the DisorderedObject if return_list is set to True.

Parameters:	sequence (str) – Amino acid sequence disorder_threshold (float) – Set to None such that it will change to 0.42 for legacy and 0.5 for metapredict. Can still manually set value. Value that defines what ‘disordered’ is based on the metapredict disorder score. The higher the value the more stringent the cutoff. Default = 0.5 for new version and 0.42 for legacy metapredict. minimum_IDR_size (int) – Defines the smallest possible IDR. This is a hard limit - i.e. we CANNOT get IDRs smaller than this. Default = 12. minimum_folded_domain (int) – Defines where we expect the limit of small folded domains to be. This is NOT a hard limit and functions to modulate the removal of large gaps (i.e. gaps less than this size are treated less strictly). Note that, in addition, gaps < 35 are evaluated with a threshold of 0.35disorder_threshold and gaps < 20 are evaluated with a threshold of 0.25disorder_threshold. These two lengthscales were decided based on the fact that coiled-coiled regions (which are IDRs in isolation) often show up with reduced apparent disorder within IDRs, and but can be as short as 20-30 residues. The folded_domain_threshold is used based on the idea that it allows a ‘shortest reasonable’ folded domain to be identified. Default=50. gap_closure (int) – Defines the largest gap that would be ‘closed’. Gaps here refer to a scenario in which you have two groups of disordered residues seprated by a ‘gap’ of un-disordered residues. In general large gap sizes will favour larger contigous IDRs. It’s worth noting that gap_closure becomes relevant only when minimum_region_size becomes very small (i.e. < 5) because really gaps emerge when the smoothed disorder fit is “noisy”, but when smoothed gaps are increasingly rare. Default=10. normalized (bool) – whether the disorder scores are normalized between zero and one, default is true return_numpy (bool) – Flag which if set to true means all numerical types are returned as numpy.ndlist. Default is True legacy (bool) – Whether to use the original metapredict network return_list (bool) – Flag that determines i to return the old format where a tuple is returned. This is retained for backwards compatibility
Returns:	By default, the function returns a DisorderObject. A DisorderObject has 7 dot variables: .sequence : str Amino acid sequence .disorder : list or np.ndaarray disorder scores .disordered_domain_boundaries : list List of domain boundaries for IDRs using Python indexing .folded_domain_boundaries : list List of domain boundaries for folded domains using Python indexing .disordered_domains : list List of the actual sequences for IDRs .folded_domains : list List of the actual sequences for folded domains
Return type:	DisorderObject
Returns:	However, if `return_list` == True. Then, the function returns a list with three elements, as outlined below. [0] - Smoothed disorder score used to aid in domain boundary identification. This can be useful for understanding how IDRs/folded domains were identified, and will vary depending on the settings provided [1] - a list of elements, where each element defines the start and end position of each IDR. If a sequence was provided the third element in each sub-element is the IDR sequence. If no sequence was provided, then each sub-element is simply len=2. [2] - a list of elements, where each element defines the start and end position of each folded region. If a sequence was provided the third element in each sub-element is the folded domain sequence. If no sequence was provided, then each sub-element is simply len=2.
Return type:	list

metapredict.meta.predict_disorder(sequence, normalized=True, return_numpy=False, legacy=False)[source]¶

Function to return disorder of a single input sequence. Returns the predicted values as a list.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. normalized (bool) – Flag which defines in the predictor should control and normalize such that all values fall between 0 and 1. The underlying learning model can, in fact output some negative values and some values greater than 1. Normalization controls for this. Default = True return_numpy (bool) – Flag which if set to true means the function returns a np.array. legacy (bool) – Whether to use the original metapredict disorder predictor.
Returns:	Returns a list of floats that corresponds to the per-residue disorder score.
Return type:	list or np.ndarray

metapredict.meta.graph_disorder(sequence, title='Predicted protein disorder', disorder_threshold=None, pLDDT_scores=False, shaded_regions=None, shaded_region_color='red', DPI=150, output_file=None, legacy=False)[source]¶

Function to plot the disorder of an input sequece. Displays immediately.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. title (str) – Sets the title of the generated figure. Default = “Predicted protein disorder” disorder_threshold (float) – Set to None by default such that if the user chooses to set legacy=True, the threshhold line will be at 0.3 and if legacy is set to false (default) then the threshold line will be at 0.5. Sets a threshold which draws a horizontal black line as a visual guide along the length of the figure. Must be a value between 0 and 1. Default = 0.3 for legacy and 0.5 for new version of metapredict. pLDDT_scores (Bool) – Sets whether to include the predicted pLDDT scores in the figure shaded_regions (list of lists) – A list of lists, where sub-elements are of length 2 and contain start and end values for regions to be shaded. Assumes that sanity checking on positions has already been done. Default is None, but if there were specific regions you wanted to highlight this might, for example, look like shaded_regions=[[1,10],[40,50]], which would shade between 1 and 10 and then between 40 and 50. This can be useful to either highlight specific IDRs or specific folded domains shaded_region_color (str or list of sts) – String that defines the color of the shaded region. The shaded region is always set with an alpha of 0.3 but the color can be any valid matplotlib color name or a hex color string (i.e. “#ff0000” is red). Alternatively a list where number of elements matches number in shaded_regions, assigning a color-per-shaded regions. DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_file (str) – If provided, the output_file variable defines the location and type of the file to be saved. This should be a file location and filename with a valid matplotlib extension (such as .png, or .pdf) and, if provided, this value is passed directly to the `matplotlib.pyplot.savefig()` function as the `fname` parameter. Default = None. legacy (bool) – whether to use the legacy metapredict predictions
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_all(sequence)[source]¶

Function to return all three types of predictions (legacy_metapredict, metapredict, and ppLDDT). Returns as a tuple of numpy arrays, with ppLDDT returned as normalized between 0 and 1 (rather than 0 and 100) so can be plotted on same axis easily.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted.
Returns:	[0] - metapredict disorder scores (updated metapredict disorder) [1] - legacy metapredict disorder (original metapredict disorder) [2] - normalized ppLDDT scores
Return type:	tuple with three np.ndarrays

metapredict.meta.percent_disorder(sequence, disorder_threshold=None, mode='threshold', legacy=False)[source]¶

Function that returns the percent disorder for any given protein. By default, uses 0.5 as a cutoff for the new version of metapredict and 0.3 for the legacy version of metapredict (values greater than or equal to 0.5 will be considered disordered). If a value for cutoff is specified, that value will be used.

Mode lets you toggle between ‘threshold’ and ‘disorder_domains’. If threshold is used a simple per-residue logic operation is applied and the fraction of residues above the disorder_threshold is used. If ‘disorder_domains’ is used then the sequence is divided into IDRs and folded domains using the predict_disordered_domains() function.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. disorder_threshold (float) – Set to None by default such that it will change depending on whether legacy is set to True or False. Sets a threshold which defines if a residue is considered disordered or not. Default for new metapredict = 0.5. Default for legacy metapredict is 0.3. mode (str) – Selector which lets you choose which mode to calculate percent disorder with. Default is ‘threshold’, meaning the percentage of disorder is calculated as what fraction of residues are above the disorder_threshold. Alternatively, ‘disorder_domains’ means we use the predict_disorder_domains() function and then calculate what fraction of the protein’s residues are in the predicted IDRs. legacy (bool) – Whether or not to use the legacy metapredict.
Returns:	Returns a floating point value between 0 and 100 that defines what percentage of the sequence is considered disordered.
Return type:	float

metapredict.meta.predict_disorder_fasta(filepath, output_file=None, normalized=True, invalid_sequence_action='convert', legacy=False)[source]¶

Function to read in a .fasta file from a specified filepath. Returns a dictionary of disorder values where the key is the fasta header and the values are the predicted disorder values.

Parameters:	filepath (str) – The path to where the .fasta file is located. The filepath should end in the file name, and can be an absolute or relative path output_file (str) – By default, a dictionary of predicted values is returned immediately. However, you can specify an output filename and path and a .csv file will be saved. This should include any file extensions. Default = None. normalized (bool) – Flag which defines in the predictor should control and normalize such that all values fall between 0 and 1. The underlying learning model can, in fact output some negative values and some values greater than 1. Normalization controls for this. Default = True invalid_sequence_action (str) – Tells the function how to deal with sequences that lack standard amino acids. Default is convert, which as the name implies converts via standard rules. See https://protfasta.readthedocs.io/en/latest/read_fasta.html for more information. legacy (bool) – Whether to use the legacy metapredict predictor. Default = False.
Returns:	If output_file is set to None (as default) then this fiction returns a dictionary of sequence ID to disorder vector. If output_file is set to a filename then a .csv file will instead be written and no return data will be provided.
Return type:	dict or None

metapredict.meta.graph_disorder_fasta(filepath, pLDDT_scores=False, disorder_threshold=None, DPI=150, output_dir=None, output_filetype='png', invalid_sequence_action='convert', indexed_filenames=False, legacy=False)[source]¶

Function to make graphs of predicted disorder from the sequences in a specified .fasta file. By default will save the generated graphs to the location output_path specified in filepath.

WARNING: It is unadvisable to not include an output directory if you are reading in a .fasta file with many sequences! This is because each graph must be closed individually before the next will appear. Therefore, you will spend a bunch of time closing each graph.

NB: You cannot specify the output file name here! By default, the file name will be the first 14 characters of the FASTA header followed by the filetype as specified by filetype. If you wish for the files to include a unique leading number (i.e. X_rest_of_name where X starts at 1 and increments) then set indexed_filenames to True. This can be useful if you have sequences where the 1st 14 characters may be identical, which would otherwise overwrite an output file.

Parameters:	filepath (str) – The path to where the .fasta file is located. The filepath should end in the file name. For example (on MacOS):filepath=”/Users/thisUser/Desktop/folder_of_seqs/interesting_proteins.fasta” pLDDT_scores (Bool) – Sets whether to include the predicted pLDDT scores from AlphaFold2 disorder_threshold (float) – Sets a threshold which draws a horizontal black line as a visual guide along the length of the figure. Must be a value between 0 and 1. DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_dir (str) – If provided, the output_dir variable defines the directory where file should besaved to be saved. This should be a writeable filepath. Default is None. Output files are saved with filename as first 14 chars of fasta header (minus bad characters) plus the appropriate file extension, as defined by filetype. output_filetype (str) – String that defines the output filetype to be used. Must be one of pdf, png, jpg. invalid_sequence_action (str) – Tells the function how to deal with sequences that lack standard amino acids. Default is convert, which as the name implies converts via standard rules. See https://protfasta.readthedocs.io/en/latest/read_fasta.html for more information. indexed_filenames (bool) – Bool which, if set to true, means filenames start with an unique integer. legacy (bool) – Whether to use the legacy metapredict predictor.
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_disorder_uniprot(uniprot_id, normalized=True, legacy=False)[source]¶

Function to return disorder of a single input sequence. Uses a Uniprot ID to get the sequence.

Parameters:	uniprot_ID (str) – The uniprot ID of the sequence to predict no_ID (str) – The uniprot ID of the sequence to predict
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.graph_disorder_uniprot(uniprot_id, title='Predicted protein disorder', pLDDT_scores=False, disorder_threshold=None, shaded_regions=None, shaded_region_color='red', DPI=150, output_file=None, legacy=False)[source]¶

Function to plot the disorder of an input sequece. Displays immediately.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. title (str) – Sets the title of the generated figure. Default = “Predicted protein disorder” pLDDT_scores (Bool) – Sets whether to include the predicted pLDDT scores from AlphaFold2 disorder_threshold (float) – Set to None by default such that it will change depending of if the user sets legacy to True of if legacy remains = False. Can still be set manually. Sets a threshold which draws a horizontal black line as a visual guide along the length of the figure. Must be a value between 0 and 1. shaded_regions (list of lists) – A list of lists, where sub-elements are of length 2 and contain start and end values for regions to be shaded. Assumes that sanity checking on positions has already been done. Default is None, but if there were specific regions you wanted to highlight this might, for example, look like shaded_regions=[[1,10],[40,50]], which would shade between 1 and 10 and then between 40 and 50. This can be useful to either highlight specific IDRs or specific folded domains shaded_region_color (str) – String that defines the color of the shaded region. The shaded region is always set with an alpha of 0.3 but the color can be any valid matplotlib color name or a hex color string (i.e. “#ff0000” is red). DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_file (str) – If provided, the output_file variable defines the location and type of the file to be saved. This should be a file location and filename with a valid matplotlib extension (such as .png, or .pdf) and, if provided, this value is passed directly to the `matplotlib.pyplot.savefig()` function as the `fname` parameter. Default = None. legacy (bool) – whether to use the legacy metapredict predictor
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_disorder_domains_uniprot(uniprot_id, disorder_threshold=None, minimum_IDR_size=12, minimum_folded_domain=50, gap_closure=10, normalized=True, return_numpy=True, legacy=False)[source]¶

This function takes an amino acid sequence, a disorder score, and returns either a DisorderObjec 4-position tuple with the information listed below.

Parameters:

uniprot_ID (String) – The uniprot ID of the sequence to predict
sequence (str) – Amino acid sequence
disorder_threshold (float) – Set to None by default such that the threshold value is is dependent on whether legacy is set to True. The default for legacy is 0.42, the default for the new metapredict is 0.5.

Value that defines what ‘disordered’ is based on the metapredict disorder score.
minimum_IDR_size (int) – Defines the smallest possible IDR. This is a hard limit - i.e. we CANNOT get IDRs smaller than this. Default = 12.
minimum_folded_domain (int) – Defines where we expect the limit of small folded domains to be. This is NOT a hard limit and functions to modulate the removal of large gaps (i.e. gaps less than this size are treated less strictly). Note that, in addition, gaps < 35 are evaluated with a threshold of 0.35*disorder_threshold and gaps < 20 are evaluated with a threshold of 0.25*disorder_threshold. These two lengthscales were decided based on the fact that coiled-coiled regions (which are IDRs in isolation) often show up with reduced apparent disorder within IDRs, and but can be as short as 20-30 residues. The folded_domain_threshold is used based on the idea that it allows a ‘shortest reasonable’ folded domain to be identified. Default=50.
gap_closure (int) – Defines the largest gap that would be ‘closed’. Gaps here refer to a scenario in which you have two groups of disordered residues seprated by a ‘gap’ of un-disordered residues. In general large gap sizes will favour larger contigous IDRs. It’s worth noting that gap_closure becomes relevant only when minimum_region_size becomes very small (i.e. < 5) because really gaps emerge when the smoothed disorder fit is “noisy”, but when smoothed gaps are increasingly rare. Default=10.
return_numpy (bool) – Flag which if set to true means all numerical types are returned as numpy.ndlist. Default is True

Returns:

Returns a DisorderObject. DisorderObject has 7 dot variables:

.sequence : str: Amino acid sequence
.disorder : list or np.ndaarray: Hybrid disorder score
.disordered_domain_boundaries : list: List of domain boundaries for IDRs using Python indexing
.folded_domain_boundaries : list: List of domain boundaries for folded domains using Python indexing
.disordered_domains : list: List of the actual sequences for IDRs
.folded_domains : list: List of the actual sequences for folded domains

Return type:

DisorderObject

metapredict.meta.predict_disorder_domains_from_external_scores(disorder, sequence=None, disorder_threshold=0.5, minimum_IDR_size=12, minimum_folded_domain=50, gap_closure=10, override_folded_domain_minsize=False, return_numpy=True)[source]¶

This function takes in disorder scores generated from another predictor and applies the same domain-decomposition algorithm as predict_disorder_domains() does to extract out congigous IDRs. For example, if one were to predict disorder using the (excellent) ODiNPred, download the resulting scores, and read the scores into a list, that list could be passed as the $disorder argument to this function.

Note that the settings used here may be inapplicable to another disorder predictor, so you may need to play around with the parameters including disorder_threshold, minimum_IDR_size, minimum_folded_domain and gap_closure.

Parameters:

disorder (list) – A list of per-residue disorder scores.
sequence (str) – The protein sequence as a string. If no sequence is passed, calling DisorderObject.sequence will return an fake sequence.
disorder_threshold (float) – Value that defines what ‘disordered’ is based on the input predictor score. The higher the value the more stringent the cutoff. Default = 0.5.
minimum_IDR_size (int) – Defines the smallest possible IDR. This is a hard limit - i.e. we CANNOT get IDRs smaller than this. Default = 12.
minimum_folded_domain (int) – Defines where we expect the limit of small folded domains to be. This is NOT a hard limit and functions to modulate the removal of large gaps (i.e. gaps less than this size are treated less strictly). Note that, in addition, gaps < 35 are evaluated with a threshold of 0.35*disorder_threshold and gaps < 20 are evaluated with a threshold of 0.25*disorder_threshold. These two lengthscales were decided based on the fact that coiled-coiled regions (which are IDRs in isolation) often show up with reduced apparent disorder within IDRs, and but can be as short as 20-30 residues. The folded_domain_threshold is used based on the idea that it allows a ‘shortest reasonable’ folded domain to be identified. Default = 50.
gap_closure (int) – Defines the largest gap that would be ‘closed’. Gaps here refer to a scenario in which you have two groups of disordered residues seprated by a ‘gap’ of un-disordered residues. In general large gap sizes will favour larger contigous IDRs. It’s worth noting that gap_closure becomes relevant only when minimum_region_size becomes very small (i.e. < 5) because really gaps emerge when the smoothed disorder fit is “noisy”, but when smoothed gaps are increasingly rare. Default = 10.
override_folded_domain_minsize (bool) – By default this function includes a fail-safe check that assumes folded domains really shouldn’t be less than 35 or 20 residues. However, for some approaches we may wish to over-ride these thresholds to match the passed minimum_folded_domain value. If this flag is set to True this override occurs. This is generally not recommended unless you expect there to be well-defined sharp boundaries which could define small (20-30) residue folded domains. This is not provided as an option in the normal predict_disorder_domains for metapredict. Default = False.
return_numpy (bool) – Flag which if set to true means all numerical types are returned as numpy.ndlist. Default is True

Returns:

Returns a DisorderObject. DisorderObject has 7 dot variables:

.sequence : str: Amino acid sequence
.disorder : list or np.ndaarray: Hybrid disorder score
.disordered_domain_boundaries : list: List of domain boundaries for IDRs using Python indexing
.folded_domain_boundaries : list: List of domain boundaries for folded domains using Python indexing
.disordered_domains : list: List of the actual sequences for IDRs
.folded_domains : list: List of the actual sequences for folded domains

Return type:

DisorderObject

metapredict.meta.graph_pLDDT_uniprot(uniprot_id, title='Predicted AF2 pLDDT Scores', shaded_regions=None, shaded_region_color='red', DPI=150, output_file=None)[source]¶

Function to plot the disorder of an input sequece. Displays immediately.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. title (str) – Sets the title of the generated figure. Default = “Predicted protein disorder” shaded_regions (list of lists) – A list of lists, where sub-elements are of length 2 and contain start and end values for regions to be shaded. Assumes that sanity checking on positions has already been done. Default is None, but if there were specific regions you wanted to highlight this might, for example, look like shaded_regions=[[1,10],[40,50]], which would shade between 1 and 10 and then between 40 and 50. This can be useful to either highlight specific IDRs or specific folded domains shaded_region_color (str) – String that defines the color of the shaded region. The shaded region is always set with an alpha of 0.3 but the color can be any valid matplotlib color name or a hex color string (i.e. “#ff0000” is red). DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_file (str) – If provided, the output_file variable defines the location and type of the file to be saved. This should be a file location and filename with a valid matplotlib extension (such as .png, or .pdf) and, if provided, this value is passed directly to the `matplotlib.pyplot.savefig()` function as the `fname` parameter. Default = None.
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_pLDDT_uniprot(uniprot_id)[source]¶

Function to return pLDDT score of a single input sequence. Uses a Uniprot ID to get the sequence.

Parameters:	uniprot_ID (str) – The uniprot ID of the sequence to predict
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.graph_pLDDT_fasta(filepath, DPI=150, output_dir=None, output_filetype='png', invalid_sequence_action='convert', indexed_filenames=False)[source]¶

Function to make graphs of predicted pLDDT from the sequences in a specified .fasta file. By default will save the generated graphs to the location output_path specified in filepath.

WARNING: It is unadvisable to not include an output directory if you are reading in a .fasta file with many sequences! This is because each graph must be closed individually before the next will appear. Therefore, you will spend a bunch of time closing each graph.

NB: You cannot specify the output file name here! By default, the file name will be the first 14 characters of the FASTA header followed by the filetype as specified by filetype. If you wish for the files to include a unique leading number (i.e. X_rest_of_name where X starts at 1 and increments) then set indexed_filenames to True. This can be useful if you have sequences where the 1st 14 characters may be identical, which would otherwise overwrite an output file.

Parameters:	filepath (str) – The path to where the .fasta file is located. The filepath should end in the file name. For example (on MacOS):filepath=”/Users/thisUser/Desktop/folder_of_seqs/interesting_proteins.fasta” DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_dir (str) – If provided, the output_dir variable defines the directory where file should besaved to be saved. This should be a writeable filepath. Default is None. Output files are saved with filename as first 14 chars of fasta header (minus bad characters) plus the appropriate file extension, as defined by filetype. output_filetype (str) – String that defines the output filetype to be used. Must be one of pdf, png, jpg. invalid_sequence_action (str) – Tells the function how to deal with sequences that lack standard amino acids. Default is convert, which as the name implies converts via standard rules. See https://protfasta.readthedocs.io/en/latest/read_fasta.html for more information. indexed_filenames (bool) – Bool which, if set to true, means filenames start with an unique integer.
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_pLDDT_fasta(filepath, output_file=None, invalid_sequence_action='convert')[source]¶

Function to read in a .fasta file from a specified filepath. Returns a dictionary of pLDDT values where the key is the fasta header and the values are the predicted pLDDT values.

Parameters:	filepath (str) – The path to where the .fasta file is located. The filepath should end in the file name, and can be an absolute or relative path

output_file : str: By default, a dictionary of predicted values is returned immediately. However, you can specify an output filename and path and a .csv file will be saved. This should include any file extensions. Default = None.
invalid_sequence_action : str: Tells the function how to deal with sequences that lack standard amino acids. Default is convert, which as the name implies converts via standard rules. See https://protfasta.readthedocs.io/en/latest/read_fasta.html for more information.

Returns:	If output_file is set to None (as default) then this fiction returns a dictionary of sequence ID to pLDDT vector. If output_file is set to a filename then a .csv file will instead be written and no return data will be provided.
Return type:	dict or None

metapredict.meta.graph_pLDDT(sequence, title='Predicted AF2 pLDDT Confidence Score', disorder_scores=False, shaded_regions=None, shaded_region_color='red', DPI=150, output_file=None)[source]¶

Function to plot the AF2 pLDDT scores of an input sequece. Displays immediately.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. title (str) – Sets the title of the generated figure. Default = “Predicted AF2 pLDDT Confidence Score” disorder_scores (Bool) – Whether to include disorder scores. Can set to False if you just want the AF2 confidence scores. Default = False shaded_regions (list of lists) – A list of lists, where sub-elements are of length 2 and contain start and end values for regions to be shaded. Assumes that sanity checking on positions has already been done. Default is None, but if there were specific regions you wanted to highlight this might, for example, look like shaded_regions=[[1,10],[40,50]], which would shade between 1 and 10 and then between 40 and 50. This can be useful to either highlight specific IDRs or specific folded domains. Default = None shaded_region_color (str) – String that defines the color of the shaded region. The shaded region is always set with an alpha of 0.3 but the color can be any valid matplotlib color name or a hex color string (i.e. “#ff0000” is red). DPI (int) – Dots-per-inch. Defines the resolution of the generated figure. Passed to the dpi argument in `matplotlib.pyplot.savefig()`. output_file (str) – If provided, the output_file variable defines the location and type of the file to be saved. This should be a file location and filename with a valid matplotlib extension (such as .png, or .pdf) and, if provided, this value is passed directly to the `matplotlib.pyplot.savefig()` function as the `fname` parameter. Default = None.
Returns:	No return object, but, the graph is saved to disk or displayed locally.
Return type:	None

metapredict.meta.predict_pLDDT(sequence, return_numpy=False, normalized=False)[source]¶

Function to return predicted pLDDT scores. pLDDT scores are the scores reported by AlphaFold2 (AF2) that provide a measure of the confidence with which AF2 has on the local structure prediction. predicted_pLDDT (ppLDDT for short) is a prediction of this confidence score generated using a LSTM-BRNN network trained on ~360,000 protein structures.

In effect, this value should be considered a prediction of how confident we are that AF2 would be able to predict the structure. This is a reasonably good proxy for the prediction that a region will be structured but is not perfect.

Parameters:	sequence (str) – Input amino acid sequence (as string) to be predicted. return_numpy (bool) – Flag which, if set to true, means the function returns a numpy array instead of a list. normalized (bool) – Flag which, if set to true, means the function returns values scaled between 0 and 1 (rather than 0 and 100).
Returns:	Returns a list (or np.ndarray) of floats that corresponds to the per-residue pLDDT score. Return type depends on the flag return_numpy
Return type:	list or np.ndarray

metapredict.meta.predict_disorder_caid(input_fasta, output_file)[source]¶

executing script for generating a caid-compliant output file for disorder predictions using a .fasta file as the input.

Parameters:	input_fasta (str) – the input file as a string that includes the file path preceeding the file name if the file is not in the curdir output_file (str) – the output file name as a string. This can include a file path to a specific save location or by default saves to the curdir
Returns:	Does not return anything, saves a file to the destination output file
Return type:	None