Person:
Biswas, Surojit

Profile Picture

Email Address

AA Acceptance Date

Birth Date

Research Projects

Organizational Units

Job Title

Last Name

Biswas

First Name

Surojit

Name

Biswas, Surojit

Filters

2017 - 201912020 - 20211

Settings

Author's Publications: search.filters.isAuthorOfPublication.819d3635-822b-403a-a439-e65a692494cc

Search Results

Now showing 1 - 2 of 2
  • Thumbnail Image
    Publication
    Tradict enables accurate prediction of eukaryotic transcriptional states from 100 marker genes
    (Nature Publishing Group, 2017) Biswas, Surojit; Kerner, Konstantin; Teixeira, Paulo José Pereira Lima; Dangl, Jeffery L.; Jojic, Vladimir; Wigge, Philip A.
    Transcript levels are a critical determinant of the proteome and hence cellular function. Because the transcriptome is an outcome of the interactions between genes and their products, it may be accurately represented by a subset of transcript abundances. We develop a method, Tradict (transcriptome predict), capable of learning and using the expression measurements of a small subset of 100 marker genes to predict transcriptome-wide gene abundances and the expression of a comprehensive, but interpretable list of transcriptional programs that represent the major biological processes and pathways of the cell. By analyzing over 23,000 publicly available RNA-Seq data sets, we show that Tradict is robust to noise and accurate. Coupled with targeted RNA sequencing, Tradict may therefore enable simultaneous transcriptome-wide screening and mechanistic investigation at large scales.
  • No Thumbnail Available
    Publication
    Low-N Protein Engineering With Data-Efficient Deep Learning
    (Nature Publishing Group, 2021-04-07) Biswas, Surojit; Khimulya, Grigory; Alley, Ethan; Esvelt, Kevin; Church, George
    Protein engineering has enormous academic and industrial potential. However, it is limited by the lack of experimental assays that are consistent with the design goal and sufficiently high-throughput to find rare, enhanced variants. Here we introduce a machine learning-guided paradigm that can use as few as 24 functionally assayed mutant sequences to build an accurate virtual fitness landscape and screen ten million sequences via in silico directed evolution. As demonstrated in two dissimilar proteins, avGFP and TEM-1 β-lactamase, top candidates from a single round are diverse and as active as engineered mutants obtained from previous high-throughput efforts. By distilling information from natural protein sequence landscapes, our model learns a latent representation of “unnaturalness”, which helps to guide search away from nonfunctional sequence neighborhoods. Subsequent low-N supervision then identifies improvements to the activity-of-interest. Taken together, our approach enables efficient use of resource intensive high-fidelity assays without sacrificing throughput, and helps to accelerate engineered proteins into the fermenter, field, and clinic.