example_predictions

Note on variant formats

Use . to indicate deletions or encode indels from the perspective of the first left unchanged base pair: e.g. CATC C for a ATC deletion. Note that depending on the source of your variants InDels might be encoded differently e.g. via -. reference can be used to indicate that no alteration is to be made to the sequence retrieved from the reference genome.

Prediction scripts arguments

run_deploy_net.py

Argument	Default	Description
--dlmodel	deepHaemWindow	Specifcy the DL model file to use e.g. .py
--batch_size	10	Batch size to process instances.
--out_dir	predictions_dir	Name or full path to directory to store the predictions.
--name_tag	eval	'Nametag to add to filenames
--do	class	Select what to do: predict the 'class' of a sequence region or the 'damage' score per class (whenn variants provided); 'damage_and_scores' will report the damages as well as the raw scores for reference and variant
--slize	all	'Comma separated list of start and end position 0.based indexed of chromatin feature classifiers (outputs) to use for the predictions e.g. '0,1,2,3,5,6' Use labels file from data preprocessing to select ids. Default 'all' will predict using all classifiers. Technically, all classifiers are used for the prediction and slize is just used to crop the output to the desired columns so this does not increase speed but saves output space.
--rounddecimals	10	Select the number of decimal places to round probability and damage scores to.
--input	None	Input region or variant file (bed/vcf) like.
--model	None	Full path to checkpoint of model to be tested.
--genome	None	Full path to refrence genome used to extract DNA sequence of the regions o interest.
--bp_context	1000	Specify number of basepairs of DNA sequence input.
--num_classes	919	Specify the number of chromatin features (output neurons).
--run_on	gpu	Select where to run on 'cpu' or 'gpu' (if available).
--gpu	0	Select device id of a single available GPU and mask the rest.

---

run_deploy_shape_combination.py

Argument	Default	Description
--dlmodel	deepHaemWindow	Specifcy the DL model file to use e.g. .py
--batch_size	1	Batch size to process instances.
--out_dir	predictions_dir	Name or full path to directory to store the predictions.
--name_tag	eval	'Nametag to add to filenames
--slize	all	'Comma separated list of start and end position 0.based indexed of chromatin feature classifiers (outputs) to use for the predictions e.g. '0,1,2,3,5,6' Use labels file from data preprocessing to select ids. Default 'all' will predict using all classifiers. Technically, all classifiers are used for the prediction and slize is just used to crop the output to the desired columns so this does not increase speed but saves output space.
--input	None	Input region or variant file (bed/vcf) like.
--model	None	Full path to checkpoint of model to be tested.
--genome	None	Full path to reference genome used to extract DNA sequence of the regions o interest.
--bp_context	1000	Specify number of basepairs of DNA sequence input.
--num_classes	919	Specify the number of chromatin features (output neurons).
--padd_ends	none	Specify if to padd with half times bp_context N's to make predictions over chromosome ends [left, right, none, both].
--add_window	0	Basepairs to add around variants of interest supplied for prediction and visualisation later.
--bin_size	1	Bin sizes (strides over the genomic window) to use running over the new sequence. Default is 1 but can be used to run predictions over windows with larger increments to reduce computation and output.
--run_on	gpu	Select where to run on 'cpu' or 'gpu' (if available).
--gpu	0	Select device id of a single available GPU and mask the rest.

---

run_deploy_seq_only.py

Argument	Default	Description
--dlmodel	deepHaemWindow	Specifcy the DL model file to use e.g. .py
--batch_size	10	Batch size to process instances.
--out_dir	predictions_dir	Name or full path to directory to store the predictions.
--name_tag	eval	'Nametag to add to filenames
--do	class	Select what to do default: predict the 'class' of a sequence or 'damage' per class or get class scores persequence (fasta file input): 'seq'.)
--slize	all	'Comma separated list of start and end position 0.based indexed of chromatin feature classifiers (outputs) to use for the predictions e.g. '0,1,2,3,5,6' Use labels file from data preprocessing to select ids. Default 'all' will predict using all classifiers. Technically, all classifiers are used for the prediction and slize is just used to crop the output to the desired columns so this does not increase speed but saves output space.
--input	None	Input region or variant file (bed/vcf) like.
--model	None	Full path to checkpoint of model to be tested.
--genome	None	Full path to reference genome used to extract DNA sequence of the regions o interest.
--bp_context	1000	Specify number of basepairs of DNA sequence input.
--num_classes	919	Specify the number of chromatin features (output neurons).
--run_on	gpu	Select where to run on 'cpu' or 'gpu' (if available).
--gpu	0	Select device id of a single available GPU and mask the rest.