Albany 2019: 20th Conversation - Abstracts

category image Albany 2019
Conversation 20
June 11-15 2019
Adenine Press (2019)

Membranes and Machine Learning: Optimizing the transport of signals and drugs across membranes

I will discuss two examples where we have used machine learning on molecular dynamics simulations and experimental data to address the complexity of cell membranes that presents a formidable challenge to conventional biophysical approaches.

Recently, we suggested a previously unknown molecular mechanism for T-cell activation based on the preferential binding and localization of the intrinsically disordered cytoplasmic signaling tails, which are easily modified by the presence of different lipid species and by the physical state of the membrane. Moreover, in a biological membrane with liquid-ordered (Lo) and liquid-disordered (Ld) domains, we show that specific lipids may play a significant role in immunoreceptor signaling and similar mechanisms may be possible in a broader context across other signaling pathways such as RAS. We introduce an unsupervised machine algorithm based on the non-negative matrix factorization combined with custom clustering for analysis of MD simulations. Specifically, we implement this algorithm to detect and describe the lateral lipid segregation in a biological membrane mimic, a ternary lipid mixture with Lo and Ld domains.

The widespread emergence of multi-drug resistance is one of the most serious barriers to effective treatment of bacterial infections in both public health and biothreat scenarios. Gram-negative bacteria in particular represent unique challenges for antibiotic design due to the combined effects of their low permeability outer membranes and their non-specific efflux pumps. Specifically, we have shown that the low permeability of the two-membrane cell envelope in P. aeruginosa and the insufficient chemical diversity of potential antibiotic compounds present a challenge to antibiotic discovery. We employ both traditional machine learning techniques and a novel, fragment-based approach to identify the key descriptors and molecular fragments within a compound governing permeability and avoidance of efflux, as well as their combined effects in the wild type. This approach allows high-throughput screening of functionally relevant fragments and offers an alternative way to search the chemical space for novel antibiotic candidates.

Rachael Mansbach 1
Cesar A López 1
Paolo Ruggerone 2
Bridget S Wilson 3
Nick Hengartner 1
Helen I Zgurskaya 4
Boian Alexandrov 1
and S. Gnanakaran 1

1Theoretical Division
Los Alamos National Laboratory
Los Alamos, NM

2Department of Physics
University of Cagliari
Cagliari, Italy

3Department of Pathology and Cancer Center
University of New Mexico
Albuquerque, NM

4Department of Chemistry and Biochemistry
University of Oklahoma
Norman, OK

Ph: (505) 500-2689
Email: gnana@lanl.gov