Partial Product Aware Machine Learning on DNA-Encoded Libraries

Abstract

DNA encoded libraries (DELs) are used for rapid large-scale screening of small molecules against a protein target. These combinatorial libraries are built through several cycles of chemistry and DNA ligation, producing large sets of DNA-tagged molecules. Training machine learning models on DEL data has been shown to be effective at predicting molecules of interest dissimilar from those in the original DEL. Machine learning chemical property prediction approaches rely on the assumption that the property of interest is linked to a single chemical structure. In the context of DNA-encoded libraries, this is equivalent to assuming that every chemical reaction fully yields the desired product. However, in practice, multistep chemical synthesis sometimes generates partial molecules. Each unique DNA tag in a DEL therefore corresponds to a set of possible molecules. Here, we leverage reaction yield data to enumerate the set of possible molecules corresponding to a given DNA tag. This paper demonstrates that training a custom GNN on this richer dataset improves accuracy and generalization performance.

A Machine-learning-based Data Analysis Method for Cell-based Selection of DNA-encoded libraries (DELs)

“…68 Thus, future work will include modeling DEL datasets with larger scale and higher chemical diversity and adapting more advanced machine learning models that can take truncated and byproducts of DELs in consideration. 51,52 In summary, we show that the approach of ECFP-based DNN model with MAP loss function can be applied to effectively process and denoise cell-based DEL selection datasets, and the method may also be suitable for other types of complex biological targets, 11 and this approach also demonstrated its potential for in silico screening of chemical libraries.…”

“…103 There are several aspects that warrants further development. First, truncated and byproducts are inevitable in DELs, 51,52,63 and they are not considered in the MAP metric or MAP model; second, CAS-DEL only contains the tripeptide scaffold and has limited chemical diversity, 81 which makes it difficult to be generalized to unknown datasets and probably has led to the relatively low recall rate in our virtual screening study; third, the framework used in the project is a traditional fully connected network, a different and more complex machine learning method may lead to better performance. 68 Thus, future work will include modeling DEL datasets with larger scale and higher chemical diversity and adapting more advanced machine learning models that can take truncated and byproducts of DELs in consideration.…”

“…In fact, selection data analysis for reliable hit picking is one of the key issues in DEL research, especially for large DELs where the library quality is compromised by the truncated and/or side products during library synthesis. [51][52][53][54][55][56] In the past, many methods have been developed to process noisy DEL selection data. [51][52][53][54][55][56][57][58][59][60][61][62][63][64][65] A commonly used technique is aggregation, which is used to reduce the variability from the relatively small number of sequencing counts.…”

“…[51][52][53][54][55][56] In the past, many methods have been developed to process noisy DEL selection data. [51][52][53][54][55][56][57][58][59][60][61][62][63][64][65] A commonly used technique is aggregation, which is used to reduce the variability from the relatively small number of sequencing counts. 56 Kuai and co-workers proposed a framework for data normalization and enrichment calculation based on the estimation of Poisson confidence interval.…”

“…Thus, machine learning is considered to be a promising approach for processing DEL datasets. 51,52,64,65 Kómár and Kalinić have reported the use of machine learning to empower the discrimination of the true potential binders from the background noise ("deldenoiser"). 51 McCloskey and coworkers trained the classification models on aggregated DEL datasets and used the models to perform virtual screening on large chemical libraries.…”