Skip to content

Commit 09f940c

Browse files
benmwebbbenmwebb
committed
Add a Google Colab notebook of the protocol
1 parent 8bee644 commit 09f940c

File tree

1 file changed

+323
-0
lines changed

1 file changed

+323
-0
lines changed

‎notebook/ModLoop.ipynb‎

Lines changed: 323 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,323 @@
1+
{
2+
"nbformat": 4,
3+
"nbformat_minor": 0,
4+
"metadata": {
5+
"colab": {
6+
"provenance": []
7+
},
8+
"kernelspec": {
9+
"name": "python3",
10+
"display_name": "Python 3"
11+
},
12+
"language_info": {
13+
"name": "python"
14+
}
15+
},
16+
"cells": [
17+
{
18+
"cell_type": "markdown",
19+
"source": [
20+
"#ModLoop\n",
21+
"\n",
22+
"ModLoop is a protocol for automated modeling of loops in protein structures. The server relies on the loop modeling routine in [MODELLER](https://salilab.org/modeller/) that predicts the loop conformations by satisfaction of spatial restraints, without relying on a database of known protein structures.\n",
23+
"\n",
24+
"The ModLoop protocol can be run in Google Colab. The first step is to install the Modeller software by running the cell below:"
25+
],
26+
"metadata": {
27+
"id": "NtLLmo9CCCOA"
28+
}
29+
},
30+
{
31+
"cell_type": "code",
32+
"execution_count": null,
33+
"metadata": {
34+
"id": "5uAWPuZzXlUq"
35+
},
36+
"outputs": [],
37+
"source": [
38+
"# Install Modeller from Sali lab website\n",
39+
"modver = \"10.8\"\n",
40+
"!wget \"https://salilab.org/modeller/{modver}/modeller_{modver}-1_amd64.deb\"\n",
41+
"!apt install \"./modeller_{modver}-1_amd64.deb\"\n",
42+
"!rm \"modeller_{modver}-1_amd64.deb\"\n",
43+
"# Add Modeller Python modules to Colab's Python path\n",
44+
"import sys\n",
45+
"sys.path.append(\"/usr/lib/python3.9/dist-packages\")"
46+
]
47+
},
48+
{
49+
"cell_type": "markdown",
50+
"source": [
51+
"Next, configure the protocol by running the following cell and giving the [MODELLER license key](https://modbase.compbio.ucsf.edu/modloop/help#modkey), uploading the [starting structure](https://modbase.compbio.ucsf.edu/modloop/help#file) in PDB or mmCIF format, and specifying the [loops to refine](https://modbase.compbio.ucsf.edu/modloop/help#loop):"
52+
],
53+
"metadata": {
54+
"id": "8G2BFSeTDSy9"
55+
}
56+
},
57+
{
58+
"cell_type": "code",
59+
"source": [
60+
"# @title\n",
61+
"import ipywidgets as widgets\n",
62+
"from ipywidgets import GridspecLayout\n",
63+
"grid = GridspecLayout(3, 2)\n",
64+
"grid[0, 0] = widgets.Label(\"Modeller license key\")\n",
65+
"grid[0, 1] = key = widgets.Text()\n",
66+
"grid[1, 0] = widgets.Label(\"Upload coordinate file\")\n",
67+
"grid[1, 1] = coord = widgets.FileUpload(multiple=False)\n",
68+
"grid[2, 0] = widgets.Label(\"Enter loop segments\")\n",
69+
"grid[2, 1] = loops = widgets.Textarea()\n",
70+
"grid"
71+
],
72+
"metadata": {
73+
"cellView": "form",
74+
"id": "-nW-3oUNYadt"
75+
},
76+
"execution_count": null,
77+
"outputs": []
78+
},
79+
{
80+
"cell_type": "markdown",
81+
"source": [
82+
"We use the input data to set up the protocol:"
83+
],
84+
"metadata": {
85+
"id": "1atrM6BBEt6o"
86+
}
87+
},
88+
{
89+
"cell_type": "code",
90+
"source": [
91+
"# Add Modeller license key\n",
92+
"with open(f\"/usr/lib/modeller{modver}/modlib/modeller/config.py\") as fh:\n",
93+
" inst_dir = fh.readline()\n",
94+
"with open(f\"/usr/lib/modeller{modver}/modlib/modeller/config.py\", \"w\") as fh:\n",
95+
" fh.write(inst_dir)\n",
96+
" fh.write(f'license = {key.value!r}\\n')\n",
97+
"\n",
98+
"# Save uploaded file to local disk\n",
99+
"in_fname = list(coord.value.keys())[0]\n",
100+
"with open(in_fname, 'wb') as fh:\n",
101+
" fh.write(coord.value[in_fname]['content'])\n",
102+
"\n",
103+
"def parse_loop_selection(loops):\n",
104+
" \"\"\"Split out loop selection and check it\"\"\"\n",
105+
" import re\n",
106+
" # capitalize and remove spaces\n",
107+
" loops = re.sub(r'\\s+', '', loops.upper())\n",
108+
" # replace null chain IDs with a single space\n",
109+
" loops = loops.replace(\"::\", \": :\")\n",
110+
"\n",
111+
" loop_data = loops.split(\":\")[:-1]\n",
112+
"\n",
113+
" # Make sure correct number of colons were given\n",
114+
" if len(loop_data) % 4 != 0:\n",
115+
" raise ValueError(\n",
116+
" \"Syntax error in loop selection: check to make sure you \"\n",
117+
" \"have colons in the correct place (there should be a \"\n",
118+
" \"multiple of 4 colons)\")\n",
119+
"\n",
120+
" total_res = 0\n",
121+
" start_res = []\n",
122+
" start_id = []\n",
123+
" end_res = []\n",
124+
" end_id = []\n",
125+
" loops = 0\n",
126+
" while loops*4+3 < len(loop_data) and loop_data[loops*4] != \"\":\n",
127+
" try:\n",
128+
" start_res.append(int(loop_data[loops*4]))\n",
129+
" end_res.append(int(loop_data[loops*4+2]))\n",
130+
" except ValueError:\n",
131+
" raise ValueError(\n",
132+
" \"Residue indices are not numeric\")\n",
133+
" start_id.append(loop_data[loops*4+1])\n",
134+
" end_id.append(loop_data[loops*4+3])\n",
135+
" # all the selected residues\n",
136+
" total_res += (end_res[-1] - start_res[-1] + 1)\n",
137+
"\n",
138+
" ################################\n",
139+
" # too long loops rejected\n",
140+
" if ((end_res[-1] - start_res[-1]) > 20\n",
141+
" or start_id[-1] != end_id[-1]\n",
142+
" or (end_res[-1] - start_res[-1]) < 0):\n",
143+
" raise ValueError(\n",
144+
" \"The loop selected is too long (>20 residues) or \"\n",
145+
" \"shorter than 1 residue or not selected properly \"\n",
146+
" \"(syntax problem?) \"\n",
147+
" \"starting position %d:%s, ending position: %d:%s\"\n",
148+
" % (start_res[-1], start_id[-1], end_res[-1], end_id[-1]))\n",
149+
" loops += 1\n",
150+
"\n",
151+
" ################################\n",
152+
" # too many or no residues rejected\n",
153+
" if total_res > 20:\n",
154+
" raise ValueError(\n",
155+
" \"Too many loop residues have been selected \"\n",
156+
" \" (selected: %d > limit:20)!\" % total_res)\n",
157+
" if total_res <= 0:\n",
158+
" raise ValueError(\n",
159+
" \"No loop residues selected!\")\n",
160+
" return loop_data\n",
161+
"\n",
162+
"def get_output_header(loop_data, nmodel):\n",
163+
" \"\"\"Return a suitable header for output model files\"\"\"\n",
164+
" residue_range = []\n",
165+
" for i in range(0, len(loop_data), 4):\n",
166+
" residue_range.append(\" %s:%s-%s:%s\" % tuple(loop_data[i:i + 4]))\n",
167+
" looplist = \"\\n\".join(residue_range)\n",
168+
" return f\"\"\"\n",
169+
"Dear User,\n",
170+
"\n",
171+
"Coordinates for the lowest energy model (out of {nmodel} sampled)\n",
172+
"are returned with the optimized loop regions, listed below:\n",
173+
"{looplist}\n",
174+
"\n",
175+
"for references please cite these two articles:\n",
176+
"\n",
177+
" A Fiser, RKG Do and A Sali,\n",
178+
" Modeling of loops in protein structures\n",
179+
" Prot. Sci. (2000) 9, 1753-1773\n",
180+
"\n",
181+
" A Fiser and A Sali,\n",
182+
" ModLoop: Automated modeling of loops in protein structures\n",
183+
" Bioinformatics. (2003) 18(19) 2500-01\n",
184+
"\n",
185+
"\n",
186+
"For further inquiries, please contact: modloop@ucsf.edu\n",
187+
"\n",
188+
"with best regards,\n",
189+
"Andras Fiser\n",
190+
"\n",
191+
"\n",
192+
"\"\"\"\n",
193+
"\n",
194+
"def add_loop_header(model, loop_data, nmodel):\n",
195+
" \"\"\"Add a header to the given model PDB or mmCIF file\"\"\"\n",
196+
" with open(model) as fin:\n",
197+
" contents = fin.read()\n",
198+
" prefix = '#' if model.endswith('.cif') else 'REMARK'\n",
199+
" with open(model, 'w') as fout:\n",
200+
" for line in get_output_header(loop_data, nmodel).split('\\n'):\n",
201+
" if line == '':\n",
202+
" fout.write(prefix + '\\n')\n",
203+
" else:\n",
204+
" fout.write(f'{prefix} {line}\\n')\n",
205+
" fout.write(contents)\n",
206+
"\n",
207+
"# Check the provided set of loop residues to refine\n",
208+
"loop_data = parse_loop_selection(loops.value)"
209+
],
210+
"metadata": {
211+
"id": "SMGFauYYZlu6"
212+
},
213+
"execution_count": null,
214+
"outputs": []
215+
},
216+
{
217+
"cell_type": "markdown",
218+
"source": [
219+
"The loop modeling protocol itself is just a short Python script that runs MODELLER on the input file uploaded earlier, selecting the loop residues given:"
220+
],
221+
"metadata": {
222+
"id": "hek0hWnsFJgR"
223+
}
224+
},
225+
{
226+
"cell_type": "code",
227+
"source": [
228+
"from modeller import Environ, Selection, ModellerError\n",
229+
"from modeller.automodel import LoopModel, refine\n",
230+
"import sys\n",
231+
"\n",
232+
"class MyLoop(LoopModel):\n",
233+
" def select_loop_atoms(self):\n",
234+
" rngs = []\n",
235+
" for i in range(0, len(loop_data), 4):\n",
236+
" rngs.append(self.residue_range(\"%s:%s\" % tuple(loop_data[i:i+2]),\n",
237+
" \"%s:%s\" % tuple(loop_data[i+2:i+4])))\n",
238+
" if len(rngs[-1]) > 30:\n",
239+
" raise ModellerError(\"loop too long\")\n",
240+
" s = Selection(rngs)\n",
241+
" if len(s.only_no_topology()) > 0:\n",
242+
" raise ModellerError(\"some selected residues have no topology\")\n",
243+
" return s\n",
244+
"\n",
245+
"def make_loop(taskid):\n",
246+
" logfile = f'{taskid}.log'\n",
247+
" print(f'Logging output to {logfile}')\n",
248+
" old_sys_stdout = sys.stdout\n",
249+
" try:\n",
250+
" sys.stdout = open(logfile, 'w')\n",
251+
" env = Environ(rand_seed=-1000-taskid)\n",
252+
" m = MyLoop(env, inimodel=in_fname, sequence='loop')\n",
253+
" if in_fname.endswith('.cif'):\n",
254+
" m.set_output_model_format('MMCIF')\n",
255+
" else:\n",
256+
" m.set_output_model_format('PDB')\n",
257+
" m.loop.md_level = refine.slow\n",
258+
" m.loop.starting_model = m.loop.ending_model = taskid\n",
259+
" m.make()\n",
260+
" return m.loop.outputs[0]\n",
261+
" finally:\n",
262+
" sys.stdout = old_sys_stdout"
263+
],
264+
"metadata": {
265+
"id": "gCfs5jsRhXPo"
266+
},
267+
"execution_count": null,
268+
"outputs": []
269+
},
270+
{
271+
"cell_type": "markdown",
272+
"source": [
273+
"Finally, we can run the protocol in parallel. The exact same protocol is run 300 times on the same inputs but with a different random seed. (This will run faster if given a CPU with more cores.) The single structure with the lowest molecular PDF (molpdf) is then selected."
274+
],
275+
"metadata": {
276+
"id": "4n58sMJMFVtL"
277+
}
278+
},
279+
{
280+
"cell_type": "code",
281+
"source": [
282+
"import multiprocessing\n",
283+
"import operator\n",
284+
"\n",
285+
"nmodel = 300\n",
286+
"with multiprocessing.Pool() as pool:\n",
287+
" best_model = min(pool.imap_unordered(make_loop, range(1, nmodel+1)),\n",
288+
" key=operator.itemgetter('molpdf'))\n",
289+
"print(f\"Best model is {best_model['name']}\")\n",
290+
"\n",
291+
"# Add an informative ModLoop header to the best model PDB/mmCIF file\n",
292+
"add_loop_header(best_model['name'], loop_data, nmodel)"
293+
],
294+
"metadata": {
295+
"collapsed": true,
296+
"id": "6lejFefP0w-H"
297+
},
298+
"execution_count": null,
299+
"outputs": []
300+
},
301+
{
302+
"cell_type": "markdown",
303+
"source": [
304+
"Run the cell below to download the selected structure:"
305+
],
306+
"metadata": {
307+
"id": "0vpeZJEBGHtu"
308+
}
309+
},
310+
{
311+
"cell_type": "code",
312+
"source": [
313+
"from google.colab import files\n",
314+
"files.download(best_model['name'])\n"
315+
],
316+
"metadata": {
317+
"id": "9_sIwcR69lww"
318+
},
319+
"execution_count": null,
320+
"outputs": []
321+
}
322+
]
323+
}

0 commit comments

Comments
 (0)