Modern Computational Tools for Chemistry with Corin Wagen

What It Covers

Corin Wagen, founder of Rowan, explains how his cloud-based quantum chemistry platform democratizes high-accuracy molecular modeling for drug designers and chemists. The platform replaces legacy Fortran-based tools requiring SSH access with a web-native interface, incorporating machine learning potentials that run 100–1,000x faster than traditional density functional theory calculations.

Key Questions Answered

• •Legacy QM tool accessibility gap: Traditional quantum chemistry packages like Gaussian, ORCA, and Q-Chem require submitting text input files via SSH to HPC servers, then manually retrieving outputs — a workflow that functions for expert computational chemists but creates prohibitive friction for experimental chemists who want occasional calculations. Rowan eliminates this by handling compute allocation, job execution, and visualization automatically through a browser interface.
• •Force field accuracy problem: Standard molecular mechanics force fields used to rank molecular conformers achieve only a Pearson correlation coefficient of ~0.4 when predicting conformer energies — barely above random. Rowan's DFT-level calculations produce accurate conformer rankings, which directly impacts decisions about molecular shape before synthesis, potentially saving a week of lab work per molecule when incorrect geometries are caught early.
• •ML potentials as DFT replacement: Machine learning interatomic potentials now achieve near-DFT accuracy at 100–1,000x lower computational cost. Rowan integrates these models, making them accessible without programming. For drug designers working with organic small molecules containing 10–15 common elements, ML potentials cover the majority of practical use cases — conformer ranking, pKa prediction, bond dissociation energies — at speeds compatible with real-time decision-making.
• •Scientific software development rule: A software engineer's estimation framework applies directly to scientific tools: getting a prototype working takes time X, making it reliable across edge cases takes 3X more, and building a usable interface takes another 3X. Academic software stops after the first seventh of total effort because publications require only the prototype. Founders building scientific software products must budget for the remaining six-sevenths that academia never completes.
• •Market creation through evangelism: Rowan's growth strategy centers on reproducing published medicinal chemistry results — such as a Novartis paper — directly within the platform to demonstrate accuracy parity with established tools. Rather than converting existing QM users, the primary target is experimental chemists unfamiliar with quantum chemistry, requiring case-by-case demonstration of specific decision points: scaffold hops, atropisomer stability, pre-synthesis reaction feasibility checks.