<div dir="ltr"><div><br></div><div>Dear <br></div><div><br></div><div>Sorry for cross-posting.<br></div><div><br></div><div>Please spread the word that we have two postdoc positions in deep learning on protein-protein interactions<br></div><div><br></div><div><b>Postdoc at Stockholm University</b><br></div><div><a href="https://www.su.se/english/about-the-university/work-at-su/available-jobs?rmpage=job&rmjob=16725&rmlang=UK" target="_blank">https://www.su.se/english/about-the-university/work-at-su/available-jobs?rmpage=job&rmjob=16725&rmlang=UK</a><br></div><div><b>Postdoc at KTH</b></div><div><div><a href="https://www.kth.se/en/om/work-at-kth/lediga-jobb/what:job/jobID:457741/where:4/" target="_blank">https://www.kth.se/en/om/work-at-kth/lediga-jobb/what:job/jobID:457741/where:4/</a>
<p class="MsoNormal"> <br></p></div><div><br></div></div><div><p><b>Project description<br></b>Protein structure is essential
for understanding their function as well as for developing drugs
targeting proteins. Recently, a deep learning method that can predict
the structure of most proteins was made freely available and a database
with predicted structure was released. However, proteins do not act
alone – they act together with other proteins. Therefore, the next major
challenge is to use these types of methods for predicting
protein-protein interactions. Initial studies from us have shown that it
is possible to predict accurate structures of a large part of dimeric
proteins using either a modified version of AlphaFold2 or
AlphaFold-multimer. However, there are still many proteins that cannot
be built accurately, nor are we able to always distinguish interacting
from non-interacting protein pairs and to build larger complexes
accurately is still an unsolved problem. In this project, we are
recruiting two postdocs to leverage recent advances in the field of
machine learning to build better deep-learning models for predicting
protein-protein interactions and to apply these methods to biologically
relevant problems.</p>
<p><i>DDLS:</i> The SciLifeLab and Wallenberg National Program for
Data-Driven Life Science (DDLS) is a 12-year initiative that focuses on
data-driven research, within fields essential for improving people´s
lives, detecting and treating diseases, protecting biodiversity and
creating sustainability. The programme will train the next generation of
life scientists and create a strong computational and data science
base. The program aims to strengthen national collaborations between
universities, bridge the research communities of life sciences and data
sciences, and create partnerships with industry, healthcare and other
national and international actors. Read more at: <a href="https://www.scilifelab.se/data-driven" rel="noopener" target="_blank">www.scilifelab.se/data-driven</a>.</p>
<p><i>Environment:</i> The Elofsson group is located at the Science
for Life Laboratory. Elofsson has worked on protein structure
predictions for more than two decades. He has worked on various
techniques, both using machine learning and other computational
techniques. His most important contributions for this work are the
methods he has developed to identify the quality of protein models,
Pcons and various versions of ProQ. The group consists currently of 5
PhD students and one senior researcher. Azizpour’s group is part of the
KTH division of Robotics, Perception and Learning. He has extensive
experience in computer vision and deep learning. The main research
directions pursued in Azizpour’s group have direct relevance to this
project which includes robustness and estimation of uncertainty,
transfer learning including knowledge distillation techniques,
non-standard deep networks e.g., graph networks and transformers, and
interpretable deep learning. Furthermore, the group has extensive
experience in deploying large experiments in GPU clusters. It consists
of 4 PhD students, 1 postdoc, and several master students/interns.</p>
<p><i>Resources:</i> The groups have access to the Berzelius computer
(funded by KAW). Berzelius is an NVIDIA® SuperPOD consisting of 60
NVIDIA® DGX-A100 compute nodes supplied by Atos. Each DGX-A100 node is
equipped with 8 NVIDIA® A100 Tensor Core GPUs, 2 AMD Epyc™ 7742 CPUs, 1
TB RAM and 15 TB of local NVMe SSD storage. The A100 GPUs have 40 GB
on-board HBM2 VRAM.</p>
<p>Selected references:</p>
<ul><li>Bryant, P, Pozzati, G. & Elofsson, A. “Improved prediction of
protein-protein interactions using AlphaFold2” bioRxiv 2021.09.15.460468
(2021) doi:10.1101/2021.09.15.460468.</li><li>David F. Burke, Patrick Bryant, ....Arne Elofsson “Towards a
structurally resolved human protein interaction network” bioRxiv
2021.11.08.467664; doi: <a href="https://doi.org/10.1101/2021.11.08.467664" target="_blank">https://doi.org/10.1101/2021.11.08.467664</a></li><li>Mehmet Akdel, .... Arne Elofsson, Tristan I Croll, Pedro Beltrao “ A
structural biology community assessment of AlphaFold 2 applications”
bioRxiv 2021.09.26.461876; doi:
<a href="https://doi.org/10.1101/2021.09.26.461876" target="_blank">https://doi.org/10.1101/2021.09.26.461876</a></li><li>Federico Baldassarre, David Menéndez Hurtado, Arne Elofsson, Hossein
Azizpour“ GraphQA: protein model quality assessment using graph
convolutional networks“ Bioinformatics, Volume 37, Issue 3, 1 February
2021, Pages 360–366, <a href="https://doi.org/10.1093/bioinformatics/btaa714" target="_blank">https://doi.org/10.1093/bioinformatics/btaa714</a></li><li>Erik Englesson, Hossein Azizpour, “Efficient Evaluation-Time
Uncertainty Estimation by Improved Distillation”, ICML 2019 Uncertainty
in Deep Learning workshop</li></ul>
<p>This is a recruitment which is part of a joint grant for two postdocs
this one at Stockholm University financed by DDLS and one at KTH
financed by WASP.</p></div><div><br></div><div><div dir="ltr" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"> <br>Yours<br><br>Arne<br><i><br></i>-----------------------------------------<br> Arne Elofsson Science for Life Laboratory<br> Tel:+46-(0)70 695 1045 Stockholm University<br> <a href="http://bioinfo.se/" target="_blank">http://bioinfo.se/</a> Box 1031, <br> Email: <a href="mailto:arne@bioinfo.se" target="_blank">arne@bioinfo.se</a> 17121 Solna, Sweden<br> Twitter: <a href="https://twitter.com/arneelof" target="_blank">https://twitter.com/arneelof</a></div><div dir="ltr"> Zoom: <span style="display:inline-flex"><a href="https://stockholmuniversity.zoom.us/my/arneelof/" target="_blank">https://stockholmuniversity.zoom.us/my/arneelof/</a><span></span></span><br> Scholar: <a href="http://scholar.google.se/citations?user=s3OCM3AAAAAJ" target="_blank">http://scholar.google.se/citations?user=s3OCM3AAAAAJ</a><br> ORCID: <a href="https://orcid.org/0000-0002-7115-9751" target="_blank">0000-0002-7115-9751</a></div></div></div></div></div></div>