Skip to content

ScFv Repeats Window Selection #119

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
eliottpark wants to merge 1 commit into
base: main
Choose a base branch
from
Open

ScFv Repeats Window Selection #119

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
54 changes: 33 additions & 21 deletions src/anarcii/input_data_processing/sequences.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -141,34 +141,42 @@ def _handle_long_sequences(self):
### Start by indentifying the minima - the sequence positions between
# two regions which the model suggests contains IG/TCR content.
minima = []
start_idx = 0
last_start = 0

# Create a copy of data for later - we will reduce the size of data as
# we iteratively search for the minima in the next 125 residues.
probs = data

# iterate through data and find minima that adhear to our conditions.
# Identify low-scoring minima to act as boundaries between IG/TCR-like
# regions in multi-domain constructs.
#
# NOTE: use indices relative to the current truncated segment
# (`local_segment.index(...)`) rather than `.index()` on the full
# probability list. For repetitive sequences (e.g. identical VH repeats),
# global `.index()` can snap to the first occurrence and make splitting
# unstable.
while len(data) > 1:
min_value = min(data[:SCFV_WINDOW_NUM])
local_segment = data[:SCFV_WINDOW_NUM]
min_value = min(local_segment)
# The minima must be global...
# And not at the end of the sequence..
# Or the start...
local_min_idx = local_segment.index(min_value)
global_min_idx = local_min_idx + last_start
if (
(min_value < SCFV_THRESHOLD)
and (
(len(probs) - (data.index(min_value) + last_start))
(len(probs) - global_min_idx)
> 10 / SCFV_JUMP
)
and (data.index(min_value) + last_start) > 10 / SCFV_JUMP
and global_min_idx > 10 / SCFV_JUMP
):
start_idx = data.index(min_value)
minima.append(start_idx + last_start)
minima.append(global_min_idx)

# We need to ensure similar minima are not too close together
# move on 5 windows (SHIFT)
data = data[start_idx + SHIFT :]
last_start += start_idx + SHIFT
data = data[local_min_idx + SHIFT :]
last_start += local_min_idx + SHIFT

# Get the original sequence.
seq = self.seqs[key]
Expand All @@ -180,20 +188,21 @@ def _handle_long_sequences(self):
idx = 1
found = 0
for i in range(len(minima)):
# For we want maxima in first 50 residues.
window = probs[offset : (offset + int(50 / SCFV_JUMP))]

# print("Offset:", offset,
# "MAX:", max(window),
# "Max IDX", probs.index(max(window)),
# "Len:", len(window))

# Shift the offset to the new minima.
offset = minima[i]
if window and max(window) > SCFV_THRESHOLD:
segment_start = offset
boundary = minima[i]
# Search the full segment between minima, not just the first ~50
# residues after a boundary.
window = probs[segment_start:boundary]

if window:
local_max = max(window)
if local_max <= SCFV_THRESHOLD:
offset = boundary
continue
# Add 2 to the peak index to ensure we contain the Ig domain
# This was simply found by trial and error (SORRY, no magic).
peak_idx_plus2 = probs.index(max(window)) + 2
local_max_idx = window.index(local_max)
peak_idx_plus2 = (segment_start + local_max_idx) + 2
new_key = f"{key}-{idx}"

# We will cap all sequences at 180.
Expand Down Expand Up @@ -235,6 +244,9 @@ def _handle_long_sequences(self):
if j < len(probs):
probs[j] = 0

# Shift the offset to the next segment boundary.
offset = boundary

if self.verbose:
print("")

Expand Down