Paul G. Francoeur


5th year Computational Biology Doctoral Candidate at CMU/PITT currently working with Dr. David Ryan Koes. I have a MS in Cell and Molecular Biology from GVSU, and a BS in the Cell and Molecular Biology and Mathematics from GVSU. My research interests lie in the development of Machine Learning algorithms and systems to aid in the drug discovery process and their utilization in the development of novel theraputics.


Department of Computational and Systems Biology
Murdoch Building
3420 Forbes Avenue
Pittsburgh, PA 15213

Download my CV

Current Research Topics

A Better Understanding of Machine Learning Models and Chemical Representations

While machine learning models have shown great success in predicting various molecular properties of interest, yet generally remain uninterpretable for explaining how a particular model is calculating the predicted property for a given input. In addition, there are numerous ways to represent compounds (graphs, grids, fingerprints, etc.) each of which contain their own advantages and disadvantages. I aim to explore the relationship between model architecture and input representation to help develop rationale for which input representations should be utilized for given tasks and help to link chemical domain knowledge to model performance.

Training Data Augmentation

Convoultional Nerual Networks are successful at image recognition, which is analagous to protein-ligand recognition, but are fundamentally limited by the amount of data that they can train on. For our purposes, the PDB is the limiting factor on the amount of training data that we can utilize. As such, exploring ways to augment the training data could prove useful in expanding the power of a CNN to effectively distinguish good and bad drug-protein pairs, while also providing a method to sidestep current scoring functions. The project, gnina, can be found here.


Mutants in Profilin have been implicated in ALS. These mutants are the current target that I am utilizing my labs methods to discover drug interactions with for a more robust knowledge of how the mutants function in the disease state and potential treatments. We have generated a novel molecule to inhibit the Profilin-Actin binding interaction.


Paul Francoeur, Daniel Penaherrera, David Ryan Koes Active Learning for Small Molecule pKa Regression; a Long Way To Go. ChemRxiv (2022).
Paul Francoeur, David Ryan Koes SolTranNet-A machine learning tool for fast aqueous solubility prediction. JCIM (2021).
Andrew McNutt, Paul Francoeur, et al. GNINA 1.0: Molecular Docking with Deep Learning JCIM (2021).
Abigail Allen, David Gau, Paul Francoeur, et al. Actin-binding protein Profilin1 promotes aggressiveness of clear cell renal cell carcinoma cells. JBC (2020).
Paul Francoeur, et al. Three-Dimensional Convolutional Neural Networks and a CrossDocked Data Set for Structure Based Drug Design. JCIM (2020).
Sunseri, J., King, J.E., Francoeur, P.G. et al. Convolutional neural network scoring and minimization in the D3R 2017 community challenge J Comput Aided Mol Des (2018).


Paul Francoeur, David Ryan Koes Regression Uncertainty Estimation for Small Molecule pKa Active Learning ACS National Meeting - San Diego (2022). Download the poster here
Paul Francoeur, David Ryan Koes Protein-Ligand Binding Affinity Prediction with GNINA 8th Drug Discovery Innovation Programme - Boston (2019). Download the poster here
Paul Francoeur, Matthew Ragoza, Rachel Rosenzweig, Jocelyn Sunseri, David Ryan Koes GNINA: Deep Learning for Molecular Docking ACS National Meeting - Boston (2018). Download the poster here

Lab Members.

PI: David Ryan Koes
PostDoc: Dakota Folmsbee
Graduate Student: Jonathan King
Graduate Student: Andrew McNutt
Graduate Student: Ian Dunn
Graduate Student: Daniel Penaherrera